Back to teamAI Engineer I

InfantSam Anton J

I'm an AI Engineer building LLM platform infrastructure, multi-agent systems, and multimodal RAG. At EverQuint, I've shipped 30+ enterprise AI agents to production.

FluentMind Platform - I build the core of FluentMind, EverQuint's AI agent platform:

→ Parallel sub-agents: a single chat spins up multiple sub-agents that work on parts of a task in parallel, coordinated by an orchestrator agent.

→ jev (model and tool routing): each sub-agent gets the right model and tools, chosen automatically for its task, with fallback when a model fails.

→ Model Council: one question goes to several LLMs, and a chair model combines their answers into one response.

→ Real-time voice: a speech-to-speech agent built on OpenAI's Realtime API.

→ Agent memory: a layered working, episodic, and semantic memory system on OpenSearch, Qdrant, and Redis.

→ Video and audio RAG: users can transcribe a video from a link or upload and ask questions about it.

→ Meeting transcription: speaker-diarized transcripts and minutes, built on NVIDIA Nemotron.

→ Evaluation: hybrid retrieval benchmarks (BM25 with dense and late-interaction retrievers and LLM reranking), plus LLM-as-judge evals that measure faithfulness and hallucination across production agents.

Associate Founder, RizeOS — I started RizeOS alongside the founder as Associate Founder and helped build the company from day one. I drove fundraising, led go-to-market strategy, and shaped the product's architecture from the ground up. As the company's sole AI engineer, I owned its entire AI roadmap: I designed, built, and shipped every AI-powered system in the product, from automation pipelines to production features. It was a single-handed mandate at the centre of the company's product and growth.

Earlier Experience — At Saipem, I built a natural language interface for Power BI with schema-constrained outputs to reduce hallucinations. At Posphene AI, I optimized real-time inference for a hybrid ResNet50-LSTM model on AWS Inferentia.