13:I[79520,["/_next/static/immutable/chunks/0yln5fuc1hcis.js","/_next/static/immutable/chunks/0wnqlfbst-58l.js","/_next/static/immutable/chunks/171oc8j8z1dhx.js"],""] 14:I[69603,["/_next/static/immutable/chunks/0yln5fuc1hcis.js","/_next/static/immutable/chunks/0wnqlfbst-58l.js","/_next/static/immutable/chunks/171oc8j8z1dhx.js"],"default"] 15:I[47420,["/_next/static/immutable/chunks/0yln5fuc1hcis.js","/_next/static/immutable/chunks/0wnqlfbst-58l.js","/_next/static/immutable/chunks/171oc8j8z1dhx.js"],"default"] 16:I[84819,["/_next/static/immutable/chunks/0yln5fuc1hcis.js","/_next/static/immutable/chunks/0wnqlfbst-58l.js","/_next/static/immutable/chunks/171oc8j8z1dhx.js"],"PortfolioProvider"] 12:T54f, (function() { try { if (typeof window !== 'undefined' && window.location.pathname === '/') { var navEntries = typeof performance !== 'undefined' && performance.getEntriesByType ? performance.getEntriesByType('navigation') : []; var navEntry = navEntries && navEntries.length > 0 ? navEntries[0] : null; var isReload = navEntry ? navEntry.type === 'reload' : false; if (isReload && window.location.hash) { window.history.replaceState(null, '', window.location.pathname); if ('scrollRestoration' in window.history) { window.history.scrollRestoration = 'manual'; } window.scrollTo(0, 0); } var onKeyDown = function(e) { if ((e.key === 'F5' || ((e.ctrlKey || e.metaKey) && (e.key === 'r' || e.key === 'R'))) && window.location.pathname === '/' && window.location.hash) { window.history.replaceState(null, '', window.location.pathname); } }; window.addEventListener('keydown', onKeyDown); } } catch (e) {} })(); 17:T9c4,

"Imagine you want to sing your favorite Japanese or Chinese anime song, but you can't read the alphabet. LyricPhonics turns complex characters into words you can actually pronounce and sing, timed perfectly to the music."

1. The Problem (Why I Built This)

Singers who want to cover international songs on TikTok, YouTube, or Reels face three big hurdles:

Unreadable Foreign Characters: Hanzi, Kanji, and Hangul cannot be pronounced without years of language study.
Academic Pinyin/Romaja Cannot Be Sung: Standard romanization lacks musical rhythm and natural pitch cues, making covers sound robotic and off-beat.
No Millisecond Timestamps: Online lyrics are just plain text without musical entry points or pause markers.

2. How I Solved It (Under The Hood)

LyricPhonics uses multimodal AI and enterprise backend engineering to make foreign lyrics immediately singable:

Clean Architecture (.NET 8): Separated into 4 independent layers (Domain, Application, Infrastructure, Api) protected by 25/25 unit tests. Any AI provider or cloud storage can be swapped without touching core business logic.
Sliding-Window Key Pool & Polly v8: Free AI APIs enforce harsh rate limits (e.g. 5 requests/minute). I engineered a sliding-window key rotator with Polly v8 exponential backoff to achieve zero HTTP 429 crashes.
Military-Grade Secret Vault: User and admin API credentials are encrypted with 100,000 PBKDF2 hashing rounds and AES-256-GCM. Plaintext keys never touch the database.
60fps Real-Time Lyric Sync: Uses an O(log N) Binary Search algorithm. It pinpoints the active lyric in under 1ms instead of lagging with slow linear scans.
Client-Side Multi-Format Export: Exports timestamped lyrics directly to LRC, SRT, and TXT subtitle files in the browser without server load.

3. Key Takeaway & Results

Reliability: 100% test coverage across core services (25/25 xUnit tests passed).
Performance: Sub-millisecond lyric synchronization running at a fluid 60fps.
Cost: $0 infrastructure cost using Cloudflare R2 and efficient sliding-window key rotation.

18:T9c4,

"Imagine you want to sing your favorite Japanese or Chinese anime song, but you can't read the alphabet. LyricPhonics turns complex characters into words you can actually pronounce and sing, timed perfectly to the music."

1. The Problem (Why I Built This)

Singers who want to cover international songs on TikTok, YouTube, or Reels face three big hurdles:

Unreadable Foreign Characters: Hanzi, Kanji, and Hangul cannot be pronounced without years of language study.
Academic Pinyin/Romaja Cannot Be Sung: Standard romanization lacks musical rhythm and natural pitch cues, making covers sound robotic and off-beat.
No Millisecond Timestamps: Online lyrics are just plain text without musical entry points or pause markers.

2. How I Solved It (Under The Hood)

LyricPhonics uses multimodal AI and enterprise backend engineering to make foreign lyrics immediately singable:

Clean Architecture (.NET 8): Separated into 4 independent layers (Domain, Application, Infrastructure, Api) protected by 25/25 unit tests. Any AI provider or cloud storage can be swapped without touching core business logic.
Sliding-Window Key Pool & Polly v8: Free AI APIs enforce harsh rate limits (e.g. 5 requests/minute). I engineered a sliding-window key rotator with Polly v8 exponential backoff to achieve zero HTTP 429 crashes.
Military-Grade Secret Vault: User and admin API credentials are encrypted with 100,000 PBKDF2 hashing rounds and AES-256-GCM. Plaintext keys never touch the database.
60fps Real-Time Lyric Sync: Uses an O(log N) Binary Search algorithm. It pinpoints the active lyric in under 1ms instead of lagging with slow linear scans.
Client-Side Multi-Format Export: Exports timestamped lyrics directly to LRC, SRT, and TXT subtitle files in the browser without server load.

3. Key Takeaway & Results

Reliability: 100% test coverage across core services (25/25 xUnit tests passed).
Performance: Sub-millisecond lyric synchronization running at a fluid 60fps.
Cost: $0 infrastructure cost using Cloudflare R2 and efficient sliding-window key rotation.

19:T9f4,

"Think of ShopOps as a flight control tower for online Etsy sellers. It calculates hidden marketplace fees so you never lose money, spots trending products, and keeps your private store data locked tight."

1. The Problem (Why I Built This)

Selling products internationally on Etsy seems simple, but hidden operational traps wipe out profits:

6 Layers of Hidden Fees: Listing fees, transaction fees, processing fees, state taxes, and mandatory 15% offsite ads mean sellers often sell hundreds of items but end up losing money.
Accidental Trademark Bans: Putting a registered word (like "Velcro" or "Barbie") in a title can instantly get a shop banned permanently.
Expensive Foreign Tools: Platforms like eRank charge $30–$99/month, yet lack automated order export formatting for popular print-on-demand fulfillment hubs.

2. How I Solved It (Under The Hood)

ShopOps combines accurate financial mathematics, fast API scraping, and browser-side cryptography:

Client-Side Zero-Knowledge Encryption: Sensitive tokens are encrypted inside the seller's browser using Web Crypto API (AES-256-GCM + PBKDF2). Even if the backend database were leaked, seller secrets remain mathematically unreadable.
Multi-Key Rotation & Rate Smoothing: Distributes background requests across multiple Etsy API keys to eliminate rate limits.
Distributed Background Worker: Heavy background jobs (competitor tracking, image rendering) run on a lightweight worker service, reducing Redis command volume by 85% via smart polling pacing.
Trademark Shield: Scans listing titles and tags against registered USPTO trademarks using word-boundary regex to prevent account suspension.
1-Click Multi-Hub Fulfillment: Formats orders instantly for Gearment, Printify, and BurgerPrints, saving sellers 2–3 hours of manual copy-pasting every day.

3. Key Takeaway & Results

Profit Clarity: 100% transparent net profit and margin breakdown down to the exact penny per order.
Security: True Zero-Knowledge architecture with 0 plaintext secrets stored on servers.
Productivity: 9 standalone intelligence tools accessible immediately without requiring store credentials.

1a:T9f4,

"Think of ShopOps as a flight control tower for online Etsy sellers. It calculates hidden marketplace fees so you never lose money, spots trending products, and keeps your private store data locked tight."

1. The Problem (Why I Built This)

Selling products internationally on Etsy seems simple, but hidden operational traps wipe out profits:

6 Layers of Hidden Fees: Listing fees, transaction fees, processing fees, state taxes, and mandatory 15% offsite ads mean sellers often sell hundreds of items but end up losing money.
Accidental Trademark Bans: Putting a registered word (like "Velcro" or "Barbie") in a title can instantly get a shop banned permanently.
Expensive Foreign Tools: Platforms like eRank charge $30–$99/month, yet lack automated order export formatting for popular print-on-demand fulfillment hubs.

2. How I Solved It (Under The Hood)

ShopOps combines accurate financial mathematics, fast API scraping, and browser-side cryptography:

Client-Side Zero-Knowledge Encryption: Sensitive tokens are encrypted inside the seller's browser using Web Crypto API (AES-256-GCM + PBKDF2). Even if the backend database were leaked, seller secrets remain mathematically unreadable.
Multi-Key Rotation & Rate Smoothing: Distributes background requests across multiple Etsy API keys to eliminate rate limits.
Distributed Background Worker: Heavy background jobs (competitor tracking, image rendering) run on a lightweight worker service, reducing Redis command volume by 85% via smart polling pacing.
Trademark Shield: Scans listing titles and tags against registered USPTO trademarks using word-boundary regex to prevent account suspension.
1-Click Multi-Hub Fulfillment: Formats orders instantly for Gearment, Printify, and BurgerPrints, saving sellers 2–3 hours of manual copy-pasting every day.

3. Key Takeaway & Results

Profit Clarity: 100% transparent net profit and margin breakdown down to the exact penny per order.
Security: True Zero-Knowledge architecture with 0 plaintext secrets stored on servers.
Productivity: 9 standalone intelligence tools accessible immediately without requiring store credentials.

1b:T87f,

"Imagine having a super-smart school assistant who read 380+ platform manuals and answers questions from students and teachers in less than a second, even when the internet is crawling."

1. The Problem (Why I Built This)

K12Online is a major national education platform, but users constantly struggle to get quick help:

Scattered Documentation: Hundreds of articles are buried in complex support menus, leading to repetitive support tickets.
Cloudflare Anti-Bot Walls: Gathering articles automatically was blocked by aggressive web application firewalls and browser fingerprint checks.
Slow & Costly Vector Databases: Hosted vector databases add noticeable network delay and monthly subscription costs for relatively compact document sets.

2. How I Solved It (Under The Hood)

I built an end-to-end data pipeline and an ultra-fast in-memory retrieval engine:

TLS/JA3 Fingerprint Spoofing: Used Python's curl_cffi to simulate genuine Chrome browser handshakes, bypassing Cloudflare security to scrape 383 guides in 3 seconds.
In-Memory Search (<1ms Latency): Tokenized documents directly in RAM using lightweight vector math, delivering instant search results without needing a paid vector database.
Pre-baked Stream Delivery: For standard step-by-step guides, the system streams pre-formatted verified answers with zero LLM token consumption, saving 90% in AI API costs.
Dual-Engine AI Failover: Features automatic failover between Groq (ultra-low latency) and Google Gemini (deep comprehension) to ensure 100% uptime.

3. Key Takeaway & Results

Speed: Sub-second response times across 380+ support articles.
Resilience: 0 Cloudflare blocks during automated ingestion runs.
Infrastructure Cost: $0 recurring vector database fees by keeping the search index in memory.

1c:T87f,

"Imagine having a super-smart school assistant who read 380+ platform manuals and answers questions from students and teachers in less than a second, even when the internet is crawling."

1. The Problem (Why I Built This)

K12Online is a major national education platform, but users constantly struggle to get quick help:

Scattered Documentation: Hundreds of articles are buried in complex support menus, leading to repetitive support tickets.
Cloudflare Anti-Bot Walls: Gathering articles automatically was blocked by aggressive web application firewalls and browser fingerprint checks.
Slow & Costly Vector Databases: Hosted vector databases add noticeable network delay and monthly subscription costs for relatively compact document sets.

2. How I Solved It (Under The Hood)

I built an end-to-end data pipeline and an ultra-fast in-memory retrieval engine:

TLS/JA3 Fingerprint Spoofing: Used Python's curl_cffi to simulate genuine Chrome browser handshakes, bypassing Cloudflare security to scrape 383 guides in 3 seconds.
In-Memory Search (<1ms Latency): Tokenized documents directly in RAM using lightweight vector math, delivering instant search results without needing a paid vector database.
Pre-baked Stream Delivery: For standard step-by-step guides, the system streams pre-formatted verified answers with zero LLM token consumption, saving 90% in AI API costs.
Dual-Engine AI Failover: Features automatic failover between Groq (ultra-low latency) and Google Gemini (deep comprehension) to ensure 100% uptime.

3. Key Takeaway & Results

Speed: Sub-second response times across 380+ support articles.
Resilience: 0 Cloudflare blocks during automated ingestion runs.
Infrastructure Cost: $0 recurring vector database fees by keeping the search index in memory.

1d:T7f2,

"An automated desktop robot that creates hundreds of marketing accounts, clicks verification emails, and organizes campaign data in minutes while the human team drinks coffee."

1. The Problem (Why I Built This)

Marketing teams at ASMLink had to manage and provision hundreds of affiliate accounts weekly:

Manual Grind: Typing account details, solving captchas, and verifying emails took 3 to 5 hours of tedious manual clicking every single day.
Frequent Bot Blocks: Standard automation scripts were quickly flagged and blocked by platform security filters.
Email Delays: Logging into webmail interfaces to click activation links caused frequent workflow bottlenecks.

2. How I Solved It (Under The Hood)

I designed a high-throughput multi-threaded Windows desktop app in C# .NET 8:

Task Parallel Library (TPL): Orchestrates up to 10 concurrent worker threads safely with cancellation tokens and thread-safe queues.
Stealth Selenium Driver: Removes the navigator.webdriver flag and introduces randomized human-like delays to bypass anti-bot detections.
Headless MailKit IMAP Client: Connects directly to mail servers over secure IMAP SSL, parses activation links from raw email bodies in milliseconds, and activates accounts without opening a browser.
Auto ChromeDriver Manager: Automatically matches the installed Google Chrome version on startup and downloads the compatible driver binary.

3. Key Takeaway & Results

Time Saved: Cut provisioning time from 4 hours down to under 5 minutes.
Accuracy: 99.5% automated registration success rate across hundreds of accounts.
Zero Manual Email Work: Completely automated inbox activation.

1e:T7f2,

"An automated desktop robot that creates hundreds of marketing accounts, clicks verification emails, and organizes campaign data in minutes while the human team drinks coffee."

1. The Problem (Why I Built This)

Marketing teams at ASMLink had to manage and provision hundreds of affiliate accounts weekly:

Manual Grind: Typing account details, solving captchas, and verifying emails took 3 to 5 hours of tedious manual clicking every single day.
Frequent Bot Blocks: Standard automation scripts were quickly flagged and blocked by platform security filters.
Email Delays: Logging into webmail interfaces to click activation links caused frequent workflow bottlenecks.

2. How I Solved It (Under The Hood)

I designed a high-throughput multi-threaded Windows desktop app in C# .NET 8:

Task Parallel Library (TPL): Orchestrates up to 10 concurrent worker threads safely with cancellation tokens and thread-safe queues.
Stealth Selenium Driver: Removes the navigator.webdriver flag and introduces randomized human-like delays to bypass anti-bot detections.
Headless MailKit IMAP Client: Connects directly to mail servers over secure IMAP SSL, parses activation links from raw email bodies in milliseconds, and activates accounts without opening a browser.
Auto ChromeDriver Manager: Automatically matches the installed Google Chrome version on startup and downloads the compatible driver binary.

3. Key Takeaway & Results

Time Saved: Cut provisioning time from 4 hours down to under 5 minutes.
Accuracy: 99.5% automated registration success rate across hundreds of accounts.
Zero Manual Email Work: Completely automated inbox activation.

1f:T923,

"A personal AI job hunter that scans 5 major tech job boards in 2 seconds, filters out fake developer jobs (like printer repair), and matches roles to your exact skillset while keeping your resume 100% private."

1. The Problem (Why I Built This)

Hunting for a software engineering job in Vietnam has several frustrating pitfalls:

Disguised "Software IT" Traps: Over 50% of listings tagged as "Developer" are actually IT helpdesk, printer maintenance, or software sales jobs.
Privacy Leaks: Uploading resumes to public job boards triggers endless spam calls from financial brokers and recruiters.
Heavy Scraping Tools: Selenium scrapers eat 1–2 GB of RAM, take minutes to run, and easily get blocked.

2. How I Solved It (Under The Hood)

I built an agentic, local-first tool powered by reverse engineering:

Zero-Browser REA (Reverse Engineering Analysis): Instead of loading heavy browsers, Job Scout VN directly extracts Next.js SSR hydration data (__NEXT_DATA__) and guest APIs, grabbing 150+ jobs in 1.5s using under 50MB RAM.
100% On-Device Privacy (Local-First): Resume parsing, skill extraction, and scoring occur entirely inside your computer's RAM. Zero candidate data ever leaves your machine.
Objective Compatibility Engine (0–100%): Automatically scores job openings: +40 points for genuine developer roles, +15 for matching tech stacks, and deducts points for unrealistic senior requirements.
Fail-Fast Facebook Gateway: Employs an isolated browser profile that executes first; if blocked, the pipeline aborts gracefully to preserve resources.
Standalone Offline Dashboard: Generates a single, self-contained HTML report with interactive salary distribution charts.

3. Key Takeaway & Results

Speed: 20x faster data collection than traditional browser scrapers.
Accuracy: 100% of disguised helpdesk postings filtered out.
Hosting Cost: $0 recurring server costs (runs locally as an Agent Skill).

20:T923,

"A personal AI job hunter that scans 5 major tech job boards in 2 seconds, filters out fake developer jobs (like printer repair), and matches roles to your exact skillset while keeping your resume 100% private."

1. The Problem (Why I Built This)

Hunting for a software engineering job in Vietnam has several frustrating pitfalls:

Disguised "Software IT" Traps: Over 50% of listings tagged as "Developer" are actually IT helpdesk, printer maintenance, or software sales jobs.
Privacy Leaks: Uploading resumes to public job boards triggers endless spam calls from financial brokers and recruiters.
Heavy Scraping Tools: Selenium scrapers eat 1–2 GB of RAM, take minutes to run, and easily get blocked.

2. How I Solved It (Under The Hood)

I built an agentic, local-first tool powered by reverse engineering:

Zero-Browser REA (Reverse Engineering Analysis): Instead of loading heavy browsers, Job Scout VN directly extracts Next.js SSR hydration data (__NEXT_DATA__) and guest APIs, grabbing 150+ jobs in 1.5s using under 50MB RAM.
100% On-Device Privacy (Local-First): Resume parsing, skill extraction, and scoring occur entirely inside your computer's RAM. Zero candidate data ever leaves your machine.
Objective Compatibility Engine (0–100%): Automatically scores job openings: +40 points for genuine developer roles, +15 for matching tech stacks, and deducts points for unrealistic senior requirements.
Fail-Fast Facebook Gateway: Employs an isolated browser profile that executes first; if blocked, the pipeline aborts gracefully to preserve resources.
Standalone Offline Dashboard: Generates a single, self-contained HTML report with interactive salary distribution charts.

3. Key Takeaway & Results

Speed: 20x faster data collection than traditional browser scrapers.
Accuracy: 100% of disguised helpdesk postings filtered out.
Hosting Cost: $0 recurring server costs (runs locally as an Agent Skill).

21:T756,

"A luxury digital showroom for Vietnam's biggest jewelry exporter, turned from design drawings into a fast, beautiful website that loads in the blink of an eye worldwide."

1. The Problem (Why I Built This)

International B2B jewelry clients expect world-class visual presentation, but the existing platform had major challenges:

Heavy High-Res Images: High-resolution jewelry photos caused long loading times and high bounce rates for overseas buyers.
Design Fidelity: Complex Figma design grids and subtle typography needed to match the brand's luxury identity with pixel-level precision.
Mobile Responsiveness: B2B partners viewing catalogs on mobile devices encountered awkward layout shifts.

2. How I Solved It (Under The Hood)

I translated Figma designs into clean, high-performance code:

Pixel-Perfect Figma to Code: Handcrafted custom responsive layouts with modern CSS and Tailwind, matching typography and spacing exactly.
Next-Gen Image Pipeline: Converted assets to lightweight WebP formats with native lazy-loading and blur placeholders, slashing page weight by 70%.
Clean WordPress Architecture: Built modular custom templates without bloated plugins, keeping page execution lean and secure.
Global Technical SEO: Configured OpenGraph tags and Schema.org structured data to ensure high visibility on Google search.

3. Key Takeaway & Results

Design Match: 100% pixel-perfect fidelity with original Figma designs.
Global Performance: Sub-second page loads for international clients across the US and Europe.

22:T756,

"A luxury digital showroom for Vietnam's biggest jewelry exporter, turned from design drawings into a fast, beautiful website that loads in the blink of an eye worldwide."

1. The Problem (Why I Built This)

International B2B jewelry clients expect world-class visual presentation, but the existing platform had major challenges:

Heavy High-Res Images: High-resolution jewelry photos caused long loading times and high bounce rates for overseas buyers.
Design Fidelity: Complex Figma design grids and subtle typography needed to match the brand's luxury identity with pixel-level precision.
Mobile Responsiveness: B2B partners viewing catalogs on mobile devices encountered awkward layout shifts.

2. How I Solved It (Under The Hood)

I translated Figma designs into clean, high-performance code:

Pixel-Perfect Figma to Code: Handcrafted custom responsive layouts with modern CSS and Tailwind, matching typography and spacing exactly.
Next-Gen Image Pipeline: Converted assets to lightweight WebP formats with native lazy-loading and blur placeholders, slashing page weight by 70%.
Clean WordPress Architecture: Built modular custom templates without bloated plugins, keeping page execution lean and secure.
Global Technical SEO: Configured OpenGraph tags and Schema.org structured data to ensure high visibility on Google search.

3. Key Takeaway & Results

Design Match: 100% pixel-perfect fidelity with original Figma designs.
Global Performance: Sub-second page loads for international clients across the US and Europe.

23:T59d,

"A high-speed data cleaning machine that gathers messy web articles, sweeps away broken HTML code, and turns them into organized files ready for AI to learn from."

1. The Problem (Why I Built This)

Raw help center articles are too messy for AI search engines:

Noisy Markup: Articles contain dynamic JavaScript buttons, broken tables, and tracking pixels that confuse AI language models.
Lack of Structure: Missing titles, undefined categories, and inconsistent formatting hurt search accuracy.

2. How I Solved It (Under The Hood)

I engineered an automated data extraction and cleaning pipeline:

High-Throughput Scraping: Automated extraction across 380+ support articles using asynchronous Python workers.
Content Sanitization: Cleaned noisy HTML into structured JSON objects containing title, breadcrumbs, tags, and clean markdown text.
Data Profiler: Checked document completeness, token lengths, and keyword frequency before feeding into AI knowledge bases.

3. Key Takeaway & Results

Efficiency: Processed 380+ platform articles in less than 3 seconds.
Data Quality: 100% structured data consistency with zero corrupted entries.

24:T59d,

"A high-speed data cleaning machine that gathers messy web articles, sweeps away broken HTML code, and turns them into organized files ready for AI to learn from."

1. The Problem (Why I Built This)

Raw help center articles are too messy for AI search engines:

Noisy Markup: Articles contain dynamic JavaScript buttons, broken tables, and tracking pixels that confuse AI language models.
Lack of Structure: Missing titles, undefined categories, and inconsistent formatting hurt search accuracy.

2. How I Solved It (Under The Hood)

I engineered an automated data extraction and cleaning pipeline:

High-Throughput Scraping: Automated extraction across 380+ support articles using asynchronous Python workers.
Content Sanitization: Cleaned noisy HTML into structured JSON objects containing title, breadcrumbs, tags, and clean markdown text.
Data Profiler: Checked document completeness, token lengths, and keyword frequency before feeding into AI knowledge bases.

3. Key Takeaway & Results

Efficiency: Processed 380+ platform articles in less than 3 seconds.
Data Quality: 100% structured data consistency with zero corrupted entries.

25:T82b,

"A smart visual diary that takes long stories from your life and turns them into a beautiful interactive graph of emotional highs and lows, showing how you grew through every milestone."

1. The Problem (Why I Built This)

Writing down memories helps people reflect, but raw text makes it hard to see the big picture:

Information Overload: People write thousands of words, but cannot easily track how their career, relationships, and mindset evolved over time.
AI Token Waste: If a user closes the browser while an AI is processing a long story, traditional servers keep running the LLM in the background, wasting costly API tokens.
Slow Processing: Parsing long memoirs (15,000+ characters) causes web requests to time out.

2. How I Solved It (Under The Hood)

I engineered a high-concurrency Go (Gin) backend paired with a smooth React timeline:

Go Goroutines with Context Cancellation: When a user disconnects or refreshes, Go's context.WithCancel instantly stops upstream AI calls, eliminating token waste.
Redis Background Task Queue: Uses BLPop queue pacing to process long memoirs asynchronously without blocking web servers.
Dual-Provider AI Engine: Groq LLaMA provides lightning-fast analysis in under 3 seconds, with automatic fallback to Gemini if rate limits are reached.
Interactive Canvas Timeline: Interactive chart with smooth 60fps zooming, scrubbing, and 1-click batch export of 34 visual milestone cards to a ZIP file.

3. Key Takeaway & Results

Performance: Analyzes 15,000-character memoirs in under 3 seconds.
Efficiency: Zero token waste on cancelled requests thanks to Go context management.
Visual Polish: Fluid 60fps interactive milestone visualization.

26:T82b,

"A smart visual diary that takes long stories from your life and turns them into a beautiful interactive graph of emotional highs and lows, showing how you grew through every milestone."

1. The Problem (Why I Built This)

Writing down memories helps people reflect, but raw text makes it hard to see the big picture:

Information Overload: People write thousands of words, but cannot easily track how their career, relationships, and mindset evolved over time.
AI Token Waste: If a user closes the browser while an AI is processing a long story, traditional servers keep running the LLM in the background, wasting costly API tokens.
Slow Processing: Parsing long memoirs (15,000+ characters) causes web requests to time out.

2. How I Solved It (Under The Hood)

I engineered a high-concurrency Go (Gin) backend paired with a smooth React timeline:

Go Goroutines with Context Cancellation: When a user disconnects or refreshes, Go's context.WithCancel instantly stops upstream AI calls, eliminating token waste.
Redis Background Task Queue: Uses BLPop queue pacing to process long memoirs asynchronously without blocking web servers.
Dual-Provider AI Engine: Groq LLaMA provides lightning-fast analysis in under 3 seconds, with automatic fallback to Gemini if rate limits are reached.
Interactive Canvas Timeline: Interactive chart with smooth 60fps zooming, scrubbing, and 1-click batch export of 34 visual milestone cards to a ZIP file.

3. Key Takeaway & Results

Performance: Analyzes 15,000-character memoirs in under 3 seconds.
Efficiency: Zero token waste on cancelled requests thanks to Go context management.
Visual Polish: Fluid 60fps interactive milestone visualization.

27:T5b1,

"A fun app that reveals secret, off-menu drinks at popular coffee shops and gives you the exact words to tell the barista so you get the perfect drink every time."

1. The Problem (Why I Built This)

Fans create amazing custom drink recipes on TikTok and Instagram, but finding and ordering them is hard:

Scattered Recipes: Secret drink recipes are spread across thousands of social media videos with no central search engine.
Ordering Anxiety: Customers don't know the proper coffee terminology to order customized drinks without confusing the barista.

2. How I Solved It (Under The Hood)

I created a modern full-stack web app optimized for fast discovery:

Fast Keyword & Ingredient Search: Instant client-side filtering by base drink, flavor syrups, and calorie ranges.
"How to Order" Script Generator: Breaks down complex recipes into clear, step-by-step phrases you can read out loud to staff.
Optimistic Community Voting: Instant bookmarking and upvoting with zero page reloads.

3. Key Takeaway & Results

Speed: Sub-50ms instant search across 100+ curated secret recipes.
Usability: Zero ordering errors with foolproof barista-friendly scripts.

28:T5b1,

"A fun app that reveals secret, off-menu drinks at popular coffee shops and gives you the exact words to tell the barista so you get the perfect drink every time."

1. The Problem (Why I Built This)

Fans create amazing custom drink recipes on TikTok and Instagram, but finding and ordering them is hard:

Scattered Recipes: Secret drink recipes are spread across thousands of social media videos with no central search engine.
Ordering Anxiety: Customers don't know the proper coffee terminology to order customized drinks without confusing the barista.

2. How I Solved It (Under The Hood)

I created a modern full-stack web app optimized for fast discovery:

Fast Keyword & Ingredient Search: Instant client-side filtering by base drink, flavor syrups, and calorie ranges.
"How to Order" Script Generator: Breaks down complex recipes into clear, step-by-step phrases you can read out loud to staff.
Optimistic Community Voting: Instant bookmarking and upvoting with zero page reloads.

3. Key Takeaway & Results

Speed: Sub-50ms instant search across 100+ curated secret recipes.
Usability: Zero ordering errors with foolproof barista-friendly scripts.

29:T700,

"An appetizing online food store that makes ordering specialty dishes as easy as picking items off a menu and paying with a quick phone scan."

1. The Problem (Why I Built This)

Food businesses lose customers when checkout is complicated:

Clunky Multi-Step Checkout: Standard e-commerce templates require too many forms, leading customers to abandon their food orders.
Slow Mobile Speeds: High-resolution food images often cause lag on mobile data connections.
Order Notification Delays: Restaurant staff missed incoming orders due to delayed email alerts.

2. How I Solved It (Under The Hood)

I developed a streamlined, mobile-first e-commerce experience:

1-Page Fast Checkout: Redesigned the checkout into a clean single-page flow with instant bank QR code and Zalo payment options.
Asset Optimization & Caching: Implemented aggressive page caching and responsive image delivery, keeping menu pages loading under 1.2 seconds.
Instant Order Notifications: Integrated automated webhooks sending instant order alerts directly to the kitchen staff's Telegram group.
Local Vietnamese SEO: Configured local business schema and product structured data to boost local Google Maps visibility.

3. Key Takeaway & Results

Conversion: Reduced checkout abandonment by over 30%.
Speed: Sub-1.2s page loads on 4G mobile networks.
Kitchen Efficiency: 0 missed customer orders through instant Telegram notifications.

2a:T700,

"An appetizing online food store that makes ordering specialty dishes as easy as picking items off a menu and paying with a quick phone scan."

1. The Problem (Why I Built This)

Food businesses lose customers when checkout is complicated:

Clunky Multi-Step Checkout: Standard e-commerce templates require too many forms, leading customers to abandon their food orders.
Slow Mobile Speeds: High-resolution food images often cause lag on mobile data connections.
Order Notification Delays: Restaurant staff missed incoming orders due to delayed email alerts.

2. How I Solved It (Under The Hood)

I developed a streamlined, mobile-first e-commerce experience:

1-Page Fast Checkout: Redesigned the checkout into a clean single-page flow with instant bank QR code and Zalo payment options.
Asset Optimization & Caching: Implemented aggressive page caching and responsive image delivery, keeping menu pages loading under 1.2 seconds.
Instant Order Notifications: Integrated automated webhooks sending instant order alerts directly to the kitchen staff's Telegram group.
Local Vietnamese SEO: Configured local business schema and product structured data to boost local Google Maps visibility.

3. Key Takeaway & Results

Conversion: Reduced checkout abandonment by over 30%.
Speed: Sub-1.2s page loads on 4G mobile networks.
Kitchen Efficiency: 0 missed customer orders through instant Telegram notifications.

2b:T5e3,

"A 3D video game for warehouse workers—you can fly around the shelves on your screen, click any box to see what's inside, and find items in seconds."

1. The Problem (Why I Built This)

Traditional warehouse software uses flat, boring 2D tables:

Hard to Locate Items: Reading a row like "Rack B, Shelf 3, Tier 4" makes it hard for workers to visualize where an item physically sits in a giant building.
Wasted Space: Managers cannot quickly spot empty slots across multi-story storage racks.

2. How I Solved It (Under The Hood)

I built an interactive 3D digital twin of a warehouse directly in the browser:

React Three Fiber & Three.js: Rendered 3D storage racks, aisles, and pallets running at a silky smooth 60fps in standard web browsers without plugins.
Color-Coded Status: Pallets change color dynamically based on occupancy, expiration dates, and priority status.
Spring Boot API Backend: Handled inventory lot updates, bin locations, and barcode lookups with solid relational database integrity.

3. Key Takeaway & Results

Visual Clarity: 60fps real-time 3D navigation on standard office laptops.
Efficiency: Drastically reduced time spent walking around searching for inventory coordinates.

2c:T5e3,

"A 3D video game for warehouse workers—you can fly around the shelves on your screen, click any box to see what's inside, and find items in seconds."

1. The Problem (Why I Built This)

Traditional warehouse software uses flat, boring 2D tables:

Hard to Locate Items: Reading a row like "Rack B, Shelf 3, Tier 4" makes it hard for workers to visualize where an item physically sits in a giant building.
Wasted Space: Managers cannot quickly spot empty slots across multi-story storage racks.

2. How I Solved It (Under The Hood)

I built an interactive 3D digital twin of a warehouse directly in the browser:

React Three Fiber & Three.js: Rendered 3D storage racks, aisles, and pallets running at a silky smooth 60fps in standard web browsers without plugins.
Color-Coded Status: Pallets change color dynamically based on occupancy, expiration dates, and priority status.
Spring Boot API Backend: Handled inventory lot updates, bin locations, and barcode lookups with solid relational database integrity.

3. Key Takeaway & Results

Visual Clarity: 60fps real-time 3D navigation on standard office laptops.
Efficiency: Drastically reduced time spent walking around searching for inventory coordinates.

2d:T5ab,

"A smart document builder that writes customized, professional job cover letters and downloads them as beautiful PDFs ready to send to employers."

1. The Problem (Why I Built This)

Writing individual cover letters for dozens of job applications is exhausting:

Formatting Frustrations: Word processors often break layout formatting, margins, and typography when converted to PDF.
Generic Letters: Job seekers struggle to tailor their experiences cleanly to specific job descriptions.

2. How I Solved It (Under The Hood)

I connected an interactive React frontend to a reliable Spring Boot backend:

Side-by-Side Live Preview: See your cover letter update in real time as you fill in company details and skills.
Server-Side PDF Generation: Uses robust Java PDF libraries to generate crisp, standard-compliant vector PDF documents with perfect margins.
ATS-Friendly Formatting: Ensures text is clean and readable by automated Applicant Tracking Systems used by recruiters.

3. Key Takeaway & Results

Speed: Generates print-ready professional PDFs in under 1 second.
Clean Architecture: Decoupled React UI and Spring Boot REST API for easy maintenance.

2e:T5ab,

"A smart document builder that writes customized, professional job cover letters and downloads them as beautiful PDFs ready to send to employers."

1. The Problem (Why I Built This)

Writing individual cover letters for dozens of job applications is exhausting:

Formatting Frustrations: Word processors often break layout formatting, margins, and typography when converted to PDF.
Generic Letters: Job seekers struggle to tailor their experiences cleanly to specific job descriptions.

2. How I Solved It (Under The Hood)

I connected an interactive React frontend to a reliable Spring Boot backend:

Side-by-Side Live Preview: See your cover letter update in real time as you fill in company details and skills.
Server-Side PDF Generation: Uses robust Java PDF libraries to generate crisp, standard-compliant vector PDF documents with perfect margins.
ATS-Friendly Formatting: Ensures text is clean and readable by automated Applicant Tracking Systems used by recruiters.

3. Key Takeaway & Results

Speed: Generates print-ready professional PDFs in under 1 second.
Clean Architecture: Decoupled React UI and Spring Boot REST API for easy maintenance.

2f:T5b3,

"A smart Google Sheets add-on that connects to Etsy, pulls product data, and calculates your exact profit so you never have to do math by hand."

1. The Problem (Why I Built This)

Small shop owners spend hours each week doing tedious math in spreadsheets:

Complex Fee Structures: Etsy charges listing fees, transaction fees, and payment fees in different currencies, confusing sellers about their real profit.
Manual Copy-Pasting: Tracking sales by manually re-typing order numbers and addresses takes up valuable product-creation time.

2. How I Solved It (Under The Hood)

I built an automated script inside Google Apps Script:

Automated Etsy API Sync: Securely fetches live shop listings and sales data directly into Google Sheets with 1 click.
Real-Time Currency Conversion: Automatically pulls live exchange rates to calculate profits in local currency.
1-Click Profit Breakdown: Calculates net profit after subtracting all platform fees, shipping costs, and supplier expenses.

3. Key Takeaway & Results

Time Saved: Saved sellers 5+ hours of manual bookkeeping every week.
Accuracy: 100% mathematical consistency with official Etsy billing statements.

30:T5b3,

"A smart Google Sheets add-on that connects to Etsy, pulls product data, and calculates your exact profit so you never have to do math by hand."

1. The Problem (Why I Built This)

Small shop owners spend hours each week doing tedious math in spreadsheets:

Complex Fee Structures: Etsy charges listing fees, transaction fees, and payment fees in different currencies, confusing sellers about their real profit.
Manual Copy-Pasting: Tracking sales by manually re-typing order numbers and addresses takes up valuable product-creation time.

2. How I Solved It (Under The Hood)

I built an automated script inside Google Apps Script:

Automated Etsy API Sync: Securely fetches live shop listings and sales data directly into Google Sheets with 1 click.
Real-Time Currency Conversion: Automatically pulls live exchange rates to calculate profits in local currency.
1-Click Profit Breakdown: Calculates net profit after subtracting all platform fees, shipping costs, and supplier expenses.

3. Key Takeaway & Results

Time Saved: Saved sellers 5+ hours of manual bookkeeping every week.
Accuracy: 100% mathematical consistency with official Etsy billing statements.

31:T56f,

"An offline safe on your computer that backs up your entire online shop—saving every photo, price, and description so your business is safe no matter what happens."

1. The Problem (Why I Built This)

Online sellers can lose everything in an instant:

Account Suspensions: An accidental policy flag or platform bug can lock sellers out of years of hard work.
Data Lock-In: Platforms make it difficult to download your original high-resolution product photos in bulk.

2. How I Solved It (Under The Hood)

I built a native C# .NET desktop application for Windows:

Automated Bulk Backup: Connects to store APIs to archive listings, product variants, inventory counts, and high-resolution images.
Portable Storage Formats: Saves data in standard JSON and CSV formats alongside full-resolution image folders.
Migration-Ready: Backed-up files can be directly uploaded to Shopify or WooCommerce within minutes.

3. Key Takeaway & Results

Speed: Archives complete stores (hundreds of listings with photos) in under 2 minutes.
Peace of Mind: 100% offline data ownership on the seller's hard drive.

32:T56f,

"An offline safe on your computer that backs up your entire online shop—saving every photo, price, and description so your business is safe no matter what happens."

1. The Problem (Why I Built This)

Online sellers can lose everything in an instant:

Account Suspensions: An accidental policy flag or platform bug can lock sellers out of years of hard work.
Data Lock-In: Platforms make it difficult to download your original high-resolution product photos in bulk.

2. How I Solved It (Under The Hood)

I built a native C# .NET desktop application for Windows:

Automated Bulk Backup: Connects to store APIs to archive listings, product variants, inventory counts, and high-resolution images.
Portable Storage Formats: Saves data in standard JSON and CSV formats alongside full-resolution image folders.
Migration-Ready: Backed-up files can be directly uploaded to Shopify or WooCommerce within minutes.

3. Key Takeaway & Results

Speed: Archives complete stores (hundreds of listings with photos) in under 2 minutes.
Peace of Mind: 100% offline data ownership on the seller's hard drive.

33:T446,

"A smart browser helper that clicks repetitive buttons in web mini-games for you, so you can earn rewards while doing your homework."

1. The Problem (Why I Built This)

Many online event mini-games require thousands of repetitive clicks to collect reward badges, wasting hours of human time.

2. How I Solved It (Under The Hood)

Lightweight Userscript (JavaScript): Runs seamlessly in the browser via Tampermonkey without external dependencies.
Human-Like Jitter Delays: Simulates natural mouse movements and random time delays to avoid simple bot detection heuristics.
Automated Game State Loop: Detects game over screens, claims reward points, and restarts game rounds automatically.

3. Key Takeaway & Results

Automation: 100% hands-free gameplay running in background browser tabs.
Safety: Built-in randomized timing to keep accounts safe from automation flags.

34:T446,

"A smart browser helper that clicks repetitive buttons in web mini-games for you, so you can earn rewards while doing your homework."

1. The Problem (Why I Built This)

Many online event mini-games require thousands of repetitive clicks to collect reward badges, wasting hours of human time.

2. How I Solved It (Under The Hood)

Lightweight Userscript (JavaScript): Runs seamlessly in the browser via Tampermonkey without external dependencies.
Human-Like Jitter Delays: Simulates natural mouse movements and random time delays to avoid simple bot detection heuristics.
Automated Game State Loop: Detects game over screens, claims reward points, and restarts game rounds automatically.

3. Key Takeaway & Results

Automation: 100% hands-free gameplay running in background browser tabs.
Safety: Built-in randomized timing to keep accounts safe from automation flags.

35:T43c,

"A digital broom that sweeps out hidden junk files, cleans up temporary clutter, and makes your Windows computer feel fast and snappy again."

1. The Problem (Why I Built This)

Over months of use, Windows computers hoard gigabytes of temporary cache files and background telemetry tasks that slow down gaming and programming performance.

2. How I Solved It (Under The Hood)

Python & PowerShell Scripts: Targets safe cleanup directories (Temp, Prefetch, Windows Update cache, DNS cache).
Safe Deletion Safeguards: Verifies file locks and prevents tampering with active system components.
1-Click Maintenance: Provides a clean menu-driven CLI interface so anyone can run system tune-ups safely.

3. Key Takeaway & Results

Disk Space: Frees up 5GB to 20GB of disk space on average.
Responsiveness: Noticeably quicker startup and lower idle memory usage.

36:T43c,

"A digital broom that sweeps out hidden junk files, cleans up temporary clutter, and makes your Windows computer feel fast and snappy again."

1. The Problem (Why I Built This)

Over months of use, Windows computers hoard gigabytes of temporary cache files and background telemetry tasks that slow down gaming and programming performance.

2. How I Solved It (Under The Hood)

Python & PowerShell Scripts: Targets safe cleanup directories (Temp, Prefetch, Windows Update cache, DNS cache).
Safe Deletion Safeguards: Verifies file locks and prevents tampering with active system components.
1-Click Maintenance: Provides a clean menu-driven CLI interface so anyone can run system tune-ups safely.

3. Key Takeaway & Results

Disk Space: Frees up 5GB to 20GB of disk space on average.
Responsiveness: Noticeably quicker startup and lower idle memory usage.

37:T417,

"A desktop app where you paste an Instagram link and instantly save the original, crystal-clear photo without blurry screenshots."

1. The Problem (Why I Built This)

Instagram disables right-click image saving, forcing users to take phone screenshots that lose clarity and crop important details.

2. How I Solved It (Under The Hood)

C# .NET WPF App: Clean desktop user interface with real-time download progress bars.
Direct CDN Resolution: Parses public media endpoints to extract the original, uncompressed photo URL directly from Instagram's content delivery network.
Zero Login Required: Works without entering your personal Instagram password, keeping your credentials secure.

3. Key Takeaway & Results

Quality: Downloads original full-resolution JPG files in under a second.
Security: 100% login-free operation.

38:T417,

"A desktop app where you paste an Instagram link and instantly save the original, crystal-clear photo without blurry screenshots."

1. The Problem (Why I Built This)

Instagram disables right-click image saving, forcing users to take phone screenshots that lose clarity and crop important details.

2. How I Solved It (Under The Hood)

C# .NET WPF App: Clean desktop user interface with real-time download progress bars.
Direct CDN Resolution: Parses public media endpoints to extract the original, uncompressed photo URL directly from Instagram's content delivery network.
Zero Login Required: Works without entering your personal Instagram password, keeping your credentials secure.

3. Key Takeaway & Results

Quality: Downloads original full-resolution JPG files in under a second.
Security: 100% login-free operation.

39:T47e,

"A 3D shooter game made entirely with Python code, where you run, jump, and shoot moving targets in a 3D arena."

1. The Problem (Why I Built This)

Learning 3D game physics, vectors, and collision detection is tough when commercial engines like Unity hide the underlying math behind complex software interfaces.

2. How I Solved It (Under The Hood)

Ursina 3D Engine & Pure Python: Modeled player movement, gravity, jumping, and camera controls using pure Object-Oriented Programming (OOP).
Raycasting Hit Detection: Implemented raycast algorithms to detect exact bullet impacts on 3D targets in real time.
Procedural Sound & Particle Effects: Triggered muzzle flashes, sound effects, and target destruction at a smooth 60fps.

3. Key Takeaway & Results

Performance: 60fps 3D gameplay running on standard laptops without dedicated GPUs.
Code Quality: Clean OOP architecture demonstrating core game math principles.

3a:T47e,

"A 3D shooter game made entirely with Python code, where you run, jump, and shoot moving targets in a 3D arena."

1. The Problem (Why I Built This)

Learning 3D game physics, vectors, and collision detection is tough when commercial engines like Unity hide the underlying math behind complex software interfaces.

2. How I Solved It (Under The Hood)

Ursina 3D Engine & Pure Python: Modeled player movement, gravity, jumping, and camera controls using pure Object-Oriented Programming (OOP).
Raycasting Hit Detection: Implemented raycast algorithms to detect exact bullet impacts on 3D targets in real time.
Procedural Sound & Particle Effects: Triggered muzzle flashes, sound effects, and target destruction at a smooth 60fps.

3. Key Takeaway & Results

Performance: 60fps 3D gameplay running on standard laptops without dedicated GPUs.
Code Quality: Clean OOP architecture demonstrating core game math principles.

3b:T454,

"An automated report printer that takes messy company numbers and turns them into neat, professional PDF bills and invoices in seconds."

1. The Problem (Why I Built This)

Companies need to generate thousands of invoices and receipts monthly, but manually formatting documents in word processors leads to ugly layout errors and hours of wasted time.

2. How I Solved It (Under The Hood)

Java & JasperReports Engine: Designed modular XML templates (JRXML) with dynamic headers, footers, and calculated subtotals.
Automated Data Binding: Mapped SQL query results directly into report tables without manual formatting.
Batch Export Pipeline: Generates crisp, standard-compliant vector PDFs ready for email delivery or printing.

3. Key Takeaway & Results

Speed: Generates hundreds of enterprise invoices in seconds.
Consistency: 100% pixel-perfect typography and page numbering.

3c:T454,

"An automated report printer that takes messy company numbers and turns them into neat, professional PDF bills and invoices in seconds."

1. The Problem (Why I Built This)

Companies need to generate thousands of invoices and receipts monthly, but manually formatting documents in word processors leads to ugly layout errors and hours of wasted time.

2. How I Solved It (Under The Hood)

Java & JasperReports Engine: Designed modular XML templates (JRXML) with dynamic headers, footers, and calculated subtotals.
Automated Data Binding: Mapped SQL query results directly into report tables without manual formatting.
Batch Export Pipeline: Generates crisp, standard-compliant vector PDFs ready for email delivery or printing.

3. Key Takeaway & Results

Speed: Generates hundreds of enterprise invoices in seconds.
Consistency: 100% pixel-perfect typography and page numbering.

3d:T4a3,

"A private mini-Netflix on your computer that automatically creates movie thumbnails, organizes videos, and plays them smoothly without uploading anything to the internet."

1. The Problem (Why I Built This)

Storing home videos and clips on a computer usually means staring at messy folders with broken thumbnails, missing durations, and unsupported file formats.

2. How I Solved It (Under The Hood)

Automated FFmpeg Pipeline: Automatically generates poster thumbnails, animated hover previews, and extracts video metadata (codec, resolution, duration).
Cinematic Dark-Mode UI: Built a fluid video browsing interface with custom controls and subtitle rendering.
100% Local Privacy: Runs entirely on your local network, ensuring personal family videos never leave your hard drive.

3. Key Takeaway & Results

Organization: Instant visual gallery generated automatically for hundreds of video clips.
Privacy: 100% offline security with zero cloud dependencies.

3e:T4a3,

"A private mini-Netflix on your computer that automatically creates movie thumbnails, organizes videos, and plays them smoothly without uploading anything to the internet."

1. The Problem (Why I Built This)

Storing home videos and clips on a computer usually means staring at messy folders with broken thumbnails, missing durations, and unsupported file formats.

2. How I Solved It (Under The Hood)

Automated FFmpeg Pipeline: Automatically generates poster thumbnails, animated hover previews, and extracts video metadata (codec, resolution, duration).
Cinematic Dark-Mode UI: Built a fluid video browsing interface with custom controls and subtitle rendering.
100% Local Privacy: Runs entirely on your local network, ensuring personal family videos never leave your hard drive.

3. Key Takeaway & Results

Organization: Instant visual gallery generated automatically for hundreds of video clips.
Privacy: 100% offline security with zero cloud dependencies.