AI Engineer · Data Engineer
I build practical AI and data systems around evaluation, data quality, observability, reliable pipelines, and evidence-backed automation.
Selected for AI Engineer / Applied AI Engineer roles in Thailand. Demos clearly distinguish deterministic fixtures from optional real-model inference.
| Project | Engineering evidence | Explore |
|---|---|---|
| Receipt AI Expense Tracker | Thai/English document extraction pipeline, financial reconciliation, human review, privacy-aware local storage; live public demo uses Groq Qwen vision with Gemini as a configured fallback, with per-IP rate limits and user-confirmed local saving. | Repo · Demo |
| Customer Support RAG Triage Agent | Bounded retrieval, typed LangGraph workflow, grounding checks, evaluation, and human-action routing; public demo uses synthetic records and deterministic generation. | Repo · Demo |
| EvalOps Lab | Retrieval evaluation, versioned evidence, regression policies, CI gates, and failure analysis. Public console displays sanitized synthetic evidence, not live model execution. | Repo · Evidence |
| VeriClaim AI | Evidence-driven claim verification, bounded agent workflow, provenance, and reproducible evaluation. Hosted demo is fixture-only. | Repo · Demo |
| OpenDQ Observatory | Operational data quality, drift monitoring, incidents, lineage, deterministic RCA, and bounded AI explanations. | Repo · Live |
| Thai Review Sentiment Intelligence | Thai NLP model training, reproducible splits, baseline comparison, macro-F1 evaluation, and FastAPI inference. | Repo · Demo |
New ML / Time-Series (synthetic-only, interactive hosted demo): Industrial Predictive Maintenance AI (Live) — causal sensor features, asset-and-time holdout, model vs heuristic baseline, and reproducible tests. These synthetic scores do not establish real-world effectiveness.
Additional data / backend engineering: Thai Procurement Intelligence · Urban Mobility Data Platform · OpsPulse AI · StockRush Go.


