manish mehta

At Centific Research, I architect controlled enterprise environments for foundation-model evaluation and RL-style post-training, including tasks, traces, verifiers, rewards, replay, guardrails, observability, and automated QA.

I've built training, preference, and evaluation workflows across a 1.8-million-person expert network and led automated data and model QA for Apple, Amazon, Microsoft, and NVIDIA. Previously, at AI Enterprise Vision, I developed fine-tuning, LoRA, quantization, and reasoning-distillation pipelines while leading conversational agents, AI coaching, knowledge graphs, and responsible-AI roadmaps.

At Stanley Black & Decker, I built and led a 30+ PhD engineering team and deployed AI and robotics across 128 plants, reducing downtime 15% and increasing throughput 20%. Earlier at Accenture Technology Labs and Disney, I delivered scalable AI systems, conversational AI, speech/NLP experiences, and developer tooling used by 500+ users.