Ankur Jain

Most CTOs talk about AI strategy. I build AI systems. Over the past 4 years at OnGraph Technologies, I've led the architecture and delivery of production AI platforms across voice, language, and machine learning — not as experiments, but as deployed systems handling real users at scale. The most significant of these is Callmira, a Voice AI agent platform I architected from the ground up. Callmira enables businesses to deploy inbound and outbound calling agents at scale — running high volume of simultaneous calls, maintaining sub-second voice latency, and extracting structured data from conversations in real time. The platform integrates with providers like ElevenLabs, OpenAI, Cartesia, Deepgram and Sarvam, giving businesses full control over their agent stack. Beyond Callmira, my team has shipped RAG-based systems across legal, education, and enterprise finance verticals — including a document intelligence platform that ingests invoices from multiple billers and delivers structured data directly into enterprise ERPs. We've also deployed on-premise LLM infrastructure using Qwen and vLLM for privacy-sensitive applications, and built an AI interviewing agent combining Tavus video AI with Pipecat pipelines. The work I'm most intellectually proud of is a reinforcement learning system we built to optimize molecular structures — applying RL to accelerate the search for viable molecules from a given initiating compound. It's the kind of problem that sits at the edge of what AI can currently do, and it's the reason I stay close to research even in a leadership role. I lead an engineering organisation of 100+ people at OnGraph, with 25 direct reports across product, platform, and AI workstreams. I completed the CTO Programme at UC Berkeley's Haas School of Business in 2022, focusing on enterprise tech strategy, AI business models, and innovation portfolio management. I