Building production AI systems — governed RAG pipelines, LLM applications, and the interfaces that make them usable
I'm an AI/ML Engineer who ships production AI systems end to end. Most recently I designed and built a multilingual HR AI assistant single-handedly — a governed RAG pipeline with citation-bound answers and guards against invented figures, wired into enterprise messaging, an HRIS and Google Workspace, with an operations console so non-engineers can manage knowledge quality themselves.
My background spans AI/ML engineering (RAG and LLM applications, OT anomaly detection, GenAI localization, content automation) and computer science fundamentals (real-time systems, game engines, C/C++) — a rare combo that lets me build things that are both technically rigorous and actually ship. Whether it's a multi-channel YouTube engine cranking out 5+ videos a day or a cloud-deployed LLM workflow, I care about making AI work in production.
Right now I'm at OlaParty/Qoretex, owning an HR AI assistant as both product manager and sole engineer — retrieval architecture, FastAPI services, a Next.js operations console, Dockerised deployment, and the testing and observability to keep it honest. My AI video pipeline work — multi-channel generation with Claude, Gemini, Kling and Veo — is currently paused.
Production AI systems, GenAI workflows, and AI-automated video pipelines. My current build is the HR AI assistant at OlaParty/Qoretex; the video pipeline work below is paused.
A fully automated AI video production system. One pipeline takes a raw concept and produces a publish-ready YouTube video — no manual steps. The stack orchestrates Claude (scripting, storyboarding), DALL-E / Kling (image & video generation), Google Veo (video synthesis), and OpenAI TTS (narration) into a single FastAPI-driven workflow.
Runs 5 channels in parallel — horror, education, kids learning, AI creatures, and meme content — each with its own genre engine, beat scheduling, caption renderer, and age-tiered output. Auto-uploads to YouTube via OAuth, sends approval previews via Telegram, and tracks state across pipeline stages to recover from partial failures.
View on GitHubAI workflows for translation adaptation, dubbing, and short-form film/video localization at Lifegame. Built avatar-based ad generation pipelines and AI-assisted script creation from app store content using GCP Cloud Workflows.
End-to-end ML pipeline for OT cybersecurity: automated OpenSearch telemetry ingestion → time-series feature engineering → anomaly detection → LLM-powered analyst summaries. Streamlit dashboard for live investigation.
Multi-agent pipeline architecture where AI agents collaborate across roles — director, implementer, reviewer. Automated coding, content generation, and workflow execution via orchestrated Claude subagents and custom tooling.
A concise, ATS-friendly resume covering my end-to-end AI/ML engineering background — from a solo-built multilingual HR AI assistant and governed RAG pipelines to production ML, cloud deployment and GenAI workflows.
Open to full-time AI/ML engineering roles, available at 1 week notice. I'd love to hear about opportunities in Singapore or remote.