Software Engineer in Bangkok • AI Systems • Multi-Agent Orchestration • PRNG/RNG Analysis • Evidence-Driven Engineering

I build with him. He ships. The pattern is consistent:

• Systems first, demos second. Automation beats manual dashboards every time.
• Every claim needs evidence: logs, metrics, reproducible scripts, and citations if we touch external data.
• Tight loops: small releases, measurable impact, fast refactors.
• Tooling is part of the product: CI/CD, eval harnesses, and observability are non-negotiable.

If he says something will ship, there will be a working artifact with telemetry.

• 🧠 AI Systems Engineering: latency/cost/quality trade-offs, deterministic interfaces, eval pipelines
• 🤖 Multi-Agent Workflows: role-based agents, tool routing, cost-aware model selection
• 🔎 PRNG/RNG Analysis: statistical testing, pattern detection, reproducible experiments
• 📈 LLM Reliability: regression tests, golden sets, canary evals, drift and cost monitoring

• Evidence-first: outputs must be testable and traceable
• Metrics matter: p95 latency, token cost, win-rate, ROI, error budgets
• Automate the boring: repeatable makefiles, cron jobs, alerts
• Ship small, measure fast: tiny PRs with observable deltas

Languages: TypeScript, Python, C#
Infra & DevOps: Docker, GitHub Actions, WSL, tmux, Make, Render/VPS
Data & ETL: Pandas, SQL, Polars, Airflow, Requests/Playwright scraping with source logging
AI/ML: PyTorch, scikit-learn, RL toolkits, Open-weight model runners, eval harnesses
Agents & Tooling: multi-agent planners, tool executors, cost routers, prompt registries

• Agent Orchestrator Starter — Role-based multi-agent framework with tool routing, retries, and budget caps. [link: unverified]
• LLM Reliability Layer — Golden-set evals, regression tests, and cost/quality dashboards. [link: unverified]
• Cost-Aware Router — Dynamic model selection by latency/cost/quality SLOs. [link: unverified]
• PRNG/RNG Toolkit — Battery of tests, visualization, and reproducible seeds for experiments. [link: unverified]
• Data Observability for LLM Apps — Tracing, prompt/version diffing, and drift alerts. [link: unverified]

If you want deeper docs, I keep design notes and metrics snapshots inside each repo.

• Reproducibility: make setup && make test && make run
• CI/CD: lint, type-check, unit tests, eval jobs, and size budgets on each PR
• Telemetry: structured logs, experiment IDs, prompt/version hashes
• Security & privacy: local first for sensitive data; redact, hash, or synthesize when needed

• 📫 Email: sitthichia.phiwon@gmail.com [verified]
• 🌐 Website/Blog: ``
• 🤝 Open to collaboration on: AI infra, agent systems, reliability, PRNG/RNG analysis

This profile layout takes inspiration from community “beautify profile” resources and open-source widgets for badges, icons, and stats.