Sayak Dutta

I’m a Senior AI/ML Engineer with 10+ years of experience, focused on building AI that ships and stays up in real production environments. I specialize in multi-agent orchestration (LangGraph, LangChain), hybrid RAG architectures, and real-time distributed systems across AWS and GCP.

My background is rooted in data engineering, entity resolution, and NLP pipelines, which gives me the infrastructure rigour many AI engineers skip. I work at the intersection of AI architecture and product delivery, partnering with product managers and business stakeholders to turn capability into outcomes.

In my current role at Siaris GmbH, I architected and shipped the company’s core multi-agent LLM platform (LangGraph + LangChain) from scratch and made it the primary production system for 100+ daily active users within 3 months. I redesigned orchestration on AWS Lambda + ECS, cutting median response latency from 12s to under 6s, and improved context relevance by 42% using LangSmith evaluation harnesses; I also reduced LLM inference costs by 45% while maintaining quality.

Across Torto Labs, Berylls, and earlier data roles, I’ve delivered enterprise-grade retrieval systems and reliable ML services—hybrid RAG at 1M+ records/month, event-driven inference pipelines reaching 99.9% uptime, and production FastAPI deployments on GCP with tight latency targets. I care deeply about evaluation, reliability, permission-scoped retrieval, PII-safe isolation, and CI/CD discipline—because great models only matter when they perform consistently under load.