Building production AI systems that work
when stakes are real.
Head of Engineering & Product with 20+ years delivering AI/ML platforms across Fortune 100 companies. From ISRO neural networks to agentic AI systems governing billions in capital decisions today.
I started my AI career at ISRO, training neural networks before transformers existed. What has stayed constant across three decades is a belief that AI systems are only as valuable as their reliability in production — not their performance on a benchmark.
Across Tech, Techstartups, Healthcare, Lifesciences, Energy & Utilities, Insurance, Banking, and Fintech I have led platform transformations that generated over $4B in measurable business outcomes. The work I am most proud of is not the technology — it is the trust that business stakeholders placed in AI systems I built, because those systems told the truth about uncertainty when it mattered.
I co-founded the CAIO Circle Tri-State Chapter to build the executive AI leadership community that I wished had existed when I was navigating these decisions alone. I publish and speak to share what the journey actually looks like from inside enterprise AI — not the demo, the production.
Three open-source platforms built from real production requirements. No demos. No prototypes.
Renewable energy interconnection cost forecasting platform at Duke Energy. Combines cost modeling, Variance narrative generation, and automated workflow orchestration for capital program decision support.
Enterprise MLOps framework for production model lifecycle management. Covers experiment tracking, model registry, automated evaluation pipelines, drift detection, and governance dashboards built for regulated industries.
Enterprise AI perception platform combining computer vision, NLP, and multimodal document understanding. Powers intelligent document processing, anomaly detection, and real-time data extraction at scale across unstructured enterprise data.
Six dimensions of production LLM health — distilled from building AI systems that govern billions in capital decisions across regulated industries.
Read the Full Article →From data infrastructure to model deployment to business translation — the complete stack of skills required to ship AI that actually works in production.
Why benchmark scores are a distraction — and the 8 measurements that will make or break your AI system when real money is on the line. Introduces the G-ARVIS framework for production LLM observability.
Single-turn accuracy metrics break down when your LLM is taking multi-step actions. What needs to change in your observability stack before you ship agentic workflows to production.
What the CloudmedAI integration at R1 RCM taught me about AI platform architecture, technical due diligence, and the engineering decisions that make or break an acquisition thesis.
Most RAG failures I have diagnosed in production have nothing to do with the retrieval algorithm. They are groundedness failures, chunking failures, and context window management failures in disguise.
Open to conversations about engineering leadership, AI platform architecture, speaking engagements, and advisory roles at companies building AI that has to work when stakes are real.