Computer Science @ NUS · SRE intern @ CPF Board · Singapore
I'm Ron, a Computer Science undergraduate at NUS. Currently an SRE intern at the Central Provident Fund Board, working across ~30 repositories on security scanning, Terraform, AKS, and Azure DevOps pipeline reliability for systems serving 4M+ CPF members. Also Vice-Chair of NUS Sheares Web, where I lead 16 developers shipping a hall intranet and React Native app used by 400+ residents.
AllIn: heads-up poker AIA poker bot you can actually play, live at allin.jianrontan.com.
- Trained a ~195K info-set CFR blueprint over 52M self-play iterations for heads-up play.
- On high-stakes river decisions it doesn't just read the blueprint: a safe, blueprint-anchored solver runs CFR live to approximate game-theory-optimal play in real time.
- Productionised end to end: Flask/gunicorn API on AWS Lightsail behind Cloudflare, with DynamoDB, ECR, and SQLite.
- CI/CD via GitHub Actions with OIDC (no long-lived AWS keys) and multi-stage Docker builds tested against a prod-equivalent image.
A job scheduler in Go on Postgres, Kafka, and Redis, built around one promise: every job runs, and no job takes effect twice. In progress.
- Microservices (HTTP API, scheduler, workers) on a 3-broker Kafka cluster (12 partitions, RF 3).
- A transactional outbox writes the job and its publish intent in one Postgres transaction, so a crash between the DB write and the Kafka publish can't lose a job.
- Next up: Redis-backed fenced leases so a stalled worker that wakes up late can't double-apply a job, plus Prometheus/Grafana SLOs and autoscaling on consumer lag, on Kubernetes behind Envoy Gateway.
Chess explanation engineEngines tell you the best move; they can't tell you why. This is a RAG pipeline that grounds LLM chess explanations in Stockfish analysis and retrieved human commentaries, so the explanation matches what the engine actually sees. Live at chess.jianrontan.com.
Languages
Infrastructure
Web & data
Where I've used all this
- CPF Board: Azure GCC infra and CI/CD: weekly security scan orchestration wiring SonarQube, Fortify, Mend, and Trivy into 20+ pipelines; Trivy IaC/image scanning in shared templates used by 170+ pipelines; Terraform for Key Vault private endpoints and AKS upgrades across non-prod and prod.
- NUS Sheares Web: GitHub Actions CI/CD with Liquibase schema migrations and a plan/apply data pipeline on Postgres; AWS infra with SST (CloudFront, Lambda SSR, S3), isolated preprod/prod on Supabase.
- Screening Eagle | Proceq: Cross-platform E2E test framework (Playwright + TypeScript, 10+ device/browser configs on BrowserStack, GitLab CI); 100+ regression cases automated with Python and AltTester.
- TES Capital: AI document automation (ChatGPT API, n8n, React, Flask): OCR + prompt workflows turning unstructured insurance data into standardised policy documents at 95% accuracy.
- Hysses: Full-stack ops platform (React, Node.js, Express, MySQL, Docker) with a custom NetSuite↔MySQL integration API.
Computer Science @ NUS · jianrontan.com