At Wysa, I lead AI architecture for a mental health platform serving 5M+ users, building production LLM systems across agentic pipelines, RAG, entity detection, and inference infrastructure.
I've delivered a hybrid PII/PHI redaction pipeline with 0.92 precision and 0.85 F1, a LangGraph-powered natural-language-to-MongoDB agent that removed 2–3 hours of engineering turnaround per request, and observability across AI inference and backend services. I also redesigned Wysa's conversational engine into modular stateless services and improved user retention by 27% through prompt and inference optimization.
Outside product engineering, I built TaleStitch to 4,000+ users and 500+ paid subscribers, contribute to Gitea, and run WDF_AI, a 235K+ follower channel teaching agentic AI, RAG, and developer tooling.
