Soumyadeep Bakshi — Co-Founder

Enterprise AI is exploding, but shipping models you actually trust is still messy. I’m building Collinear to fix that. We stress-test LLMs with automated red-teams, judge model responses like domain experts, then curate high signal synthetic data that closes real performance gaps. Before Collinear, I led enterprise AI & GTM strategy at Deloitte, working with Fortune 500 clients and leading model builders on AI trust and safety. I also co-founded Kaizen.ai, a GenAI platform for buy-side research, where I learned the speed and ambiguity of zero-to-one. Collinear was born from that conviction. Our platform assesses, red-teams, and improves large language models (LLMs) before they hit production. We’ve built proprietary “judges” that emulate expert human evaluators and let teams evaluate models at scale with precision. We don’t just flag hallucinations or harmful outputs, we help fix them with targeted, synthetic training data to close performance gaps. If you’re deploying AI in regulated or reputation-sensitive environments, let’s talk. I write here about model improvement, founder decision-making, and turning demos into dependable products.

Stackforce AI infers this person is a leader in AI solutions for enterprise applications.

Location: New York, New York, United States

Experience: 10 yrs 9 mos