Patcharaphon Thepwijit

AVAILABLE NOW

Patcharaphon Thepwijit

Software Engineer

2+ Years Experience

I utilize AI to move faster on real business problems
while keeping decisions grounded in logic and reliability.

Location Bangkok, Thailand 🇹🇭

Connect
Download Resume

Experience

Cyber Defense Co., Ltd. Oct 2025 - Present

Software Engineer Intern (Fullstack)

  • Worked in agile development environments and adapted to evolving requirements.
  • Designed and implemented a software licensing platform for licensing and usage tracking.
  • Integrated license validation and usage limits into the application.

Monster Connect Co.,Ltd. Oct 2025 - Present

Software Engineer Intern (Fullstack)

  • Improved and developed AI models for an e-commerce customer support chatbot.
  • Designed and implemented an n8n workflow to automate blog content generation and publishing to CMS, Facebook, and LinkedIn.
  • Developed a LINE bot using n8n to automate customer interactions and service workflows.

Distributed Web Crawler System

Featured
Source Code
Go Redis Docker
  • Built a Redis-backed crawler where workers consume URLs via BLPOP and fan out concurrent fetchers using Go goroutines.
  • Implemented crawl correctness controls with SADD URL deduplication, MAX_DEPTH scheduling, and per-domain throttling using Redis SETNX + expiry keys.
  • Extracted title, description, and outgoing links into worker-scoped JSONL outputs, making the dataset immediately usable for indexing and analytics pipelines.

Redis-Like Key-Value Database Engine

Source Code
Go TCP
  • Implemented a TCP command server on port 6379 supporting PING, SET, GET, DEL, EXPIRE, TTL, and SAVE with line-based protocol parsing.
  • Designed a mutex-protected in-memory store plus a background expiration sweeper running every second for deterministic TTL behavior.
  • Added JSON snapshot persistence (`dump.json`) with startup restore and expiration metadata, so state survives restarts without external dependencies.

Asynchronous Job Queue & Worker System

Source Code
Node.js Redis
  • Built a producer-worker pipeline where jobs are modeled as structured JSON and consumed via BRPOP with high-priority queue preference.
  • Implemented delayed execution using a Redis sorted set and a scheduler loop that promotes due jobs to active queues at fixed intervals.
  • Added resilience controls with retry counting, dead-letter routing, worker concurrency tuning, and horizontal scaling via Docker Compose.

High-Performance Reverse Proxy Load Balancer

Source Code
Go HTTP/TCP
  • Implemented a standard-library reverse proxy with atomic round-robin backend selection that is safe under concurrent request load.
  • Built active health checks every 10 seconds (`GET /health`) and skip-on-failure routing so unhealthy nodes are automatically removed from rotation.
  • Added retry/failover logic in ReverseProxy ErrorHandler to re-route requests before response bytes are written, preventing partial client responses.

Distributed API Rate Limiting Middleware

Source Code
Python FastAPI Redis
  • Designed middleware-driven rate limiting with pluggable Token Bucket and Sliding Window algorithms backed by shared Redis state across app instances.
  • Implemented standards-friendly HTTP 429 handling with `X-RateLimit-*` and `Retry-After` headers so clients can throttle correctly instead of blindly retrying.
  • Validated stability under k6 load at ~373 req/s (22,422 requests, p95 16.51ms, 0% failures), confirming predictable behavior under burst traffic.

Factorial.ai: AI-Powered Fake News Detector

Source Code
Next.js 15 TypeScript Gemini API Neon Postgres
  • Built an AI-powered misinformation analysis platform that evaluates user text/URLs via Gemini and returns explanation + confidence signals.
  • Implemented end-to-end product flows with JWT authentication, history tracking, CSV export, and a forum/feedback loop for collaborative verification.
  • Engineered a serverless-ready architecture using Next.js API routes + Neon Postgres + NewsAPI for real-time feeds and deployability on Vercel.

Predictive Modeling of Student Performance

View Document
Decision Trees K-Means
  • Predictive model with 70-75% accuracy.
  • Extracted key behavioral drivers.

Smart Parking System: Business Analysis & Design

View Document
UML ERD WBS
  • Designed use-case/activity/class/ER diagrams.
  • Managed 90-day timeline with critical path.