Case Studies

Selected Work

Problem → Solution → Impact.
Every system is engineered to a defensible standard.

01/Open Source SecurityMerged · PR #1400

PyThaiNLP

PythonSecurityCodeQL

Description

Fixed a CodeQL-flagged ReDoS vulnerability in PyThaiNLP’s ULMFiT URL parser.

Core Achievement

Resolved a regular-expression denial-of-service vulnerability in the ULMFiT preprocessing pipeline (PR #1400).

Technical Edge

Removed nested greedy quantification from URL matching to eliminate catastrophic backtracking while preserving the intended matching behavior.

Metric

Linear-time worst-case URL matching after the security fix.

security
02/Systems & Internals

Mini-Redis

C · POSIXTCP/IPRESP Protocol

Description

High-performance in-memory key-value store from scratch in C, modeled after Redis internals.

Core Achievement

Engineered a high-performance, in-memory key-value store from scratch in C, modeled after the Redis internal architecture.

Technical Edge

Implemented an event-driven concurrency model using poll() for I/O multiplexing, eliminating thread context-switching overhead. Designed a custom hash map with dynamic resizing and guaranteed data durability through an Append-Only File (AOF) persistence engine utilizing atomic fsync syscalls.

Metric

O(1) average lookup complexity with zero-busy-wait CPU overhead.

memory
03/Mission-Critical Backend

TPSystem

Spring BootPostgreSQLRedis

Description

Banking transaction engine designed to solve complex concurrency failures and ensure absolute data integrity.

Core Achievement

A transaction processing engine designed to solve complex concurrency failures and ensure absolute data integrity.

Technical Edge

Eliminated race conditions using ACID transactions and pessimistic locking with strict resource ordering to prevent deadlocks. Built a tamper-evident audit trail using cryptographic hash-chaining and implemented idempotency patterns to ensure consistency across network retries.

Metric

100% Transaction Accuracy under high-concurrency load testing.

account_balance
04/Infrastructure & Product

AgriscanPro

DjangoReact · TSAWS

Description

End-to-end research management platform for agricultural data, built for real-world laboratory workflows.

Core Achievement

An end-to-end research management platform for agricultural data, built for real-world laboratory workflows.

Technical Edge

Architected a secure Hierarchical RBAC system across 6 distinct roles. Optimized high-volume data exports via server-side stream processing, reducing browser memory pressure by 70%. Orchestrated a cloud-native deployment using AWS (S3/CloudFront) with automated CI/CD pipelines.

Metric

Deployed for real-world research use at Thammasat University.

biotech
05/AI & RAG Systems

SPEX-Shop

Live Demoopen_in_new
Node.js · TSNext.jsMongoDB Atlas

Description

Thai smartphone spec assistant powered by RAG — scrapes real spec data and answers natural language queries in Thai.

Core Achievement

Built a full-stack RAG pipeline that scrapes Specphone.com, stores specs with vector embeddings in MongoDB Atlas, and answers Thai-language queries via a Next.js chat UI.

Technical Edge

Implemented multi-query expansion with Typhoon LLM and a neural Cross-Encoder reranker (ms-marco-MiniLM-L-6-v2) for high-precision retrieval. Engineered backend-first chat persistence with fire-and-forget DB writes, keeping response latency unaffected. Deployed on Azure App Service (backend) and Vercel (frontend) with automated monthly re-scraping via GitHub Actions.

Metric

Full RAG pipeline with neural reranking deployed to production.

smart_toy
06/Infrastructure & Observability

AgriScan Monitor

Next.js · TSAWSGrafana · OpenTelemetry

Description

Production infrastructure dashboard and real-time control panel for AgriScan Pro’s AWS services.

Core Achievement

Built an infrastructure dashboard for RDS, Elastic Beanstalk, and ElastiCache with Google OAuth access control.

Technical Edge

Enforced least-privilege IAM with an email allowlist for Server Actions. Wired Grafana and CloudWatch for P90/P99 latency, CPU saturation, memory, and FinOps cost tracking, then instrumented the system with OpenTelemetry.

Metric

Production AWS monitoring with real-time control and health smoke tests.

monitoring