Hi, I'm Sai Teja — an AI/ML engineer.
Strategic AI/ML engineer designing end-to-end intelligent systems that transform raw data into autonomous actions — multi-agent AI, RAG pipelines, and production ML infrastructure at scale.
I design and ship automation and intelligent workflows for regulated, high-stakes domains — multi-agent systems, RAG at scale, solid data modeling, and the evaluation and MLOps practices that make models trustworthy in production.
I care most about the unglamorous layer around the model: metrics you can trust, guardrails in sensitive settings, and architecture that still makes sense a year later.
When I experiment on my own time, it's often end-to-end language-model work — for example a compact decoder-only LM trained with a frugal, laptop-friendly pipeline.
Grit & adaptability
I trained a 50M-param language model from scratch on a Mac. I question the default that big problems need big budgets.
Evaluation-first
Every LLM feature ships with a Quality Bar — RAGAS, G-Eval, and HITL loops. If we can't measure it, we can't trust it.
Strategic leadership
I guide cross-functional teams through dimensional modeling and model-optimization challenges — clarity beats cleverness.
Clean architecture
Clear boundaries between data, domain, and infrastructure keep AI systems maintainable long after launch.