At Symplix AI, I build FastAPI systems that ingest tens-of-GB source packs and power real-time AI agents for up to 10K concurrent sessions. I also designed event-sourced session infrastructure handling 100K events per minute for ordered replay and fan-out.
I built durable job execution with database-backed claims, reconciliation, dead-peer detection, and at-least-once webhook delivery. I also authored a typed async Python SDK with circuit breaking, retries, backpressure, and idempotency support.
Previously at IVA AI, I built LLM-powered support and sales agents. My RAG pipeline automated roughly 60% of manual support and reduced fine-tuning costs by about $300 per tenant, while the sales agent automated roughly 90% of the sales lifecycle.
I enjoy building resilient real-time platforms, including WebRTC discussions supporting 10K+ concurrent sessions and under-500ms audio latency, as well as durable workflow engines for backend and AI tasks.
