Preetam Mukherjee

Product and engineering leader. 15+ years at the intersection of computer vision, video infrastructure, and the physical edge.

At Mahamaia Systems, I'm building OUTPOST — a neural video codec that extracts semantic features from sensor data before transmission, enabling perception over bandwidth-constrained links like drones and subsea vehicles. I built the full encoder-decoder pipeline in Python, designed eval frameworks, and built agentic AI systems with LangGraph and RAG pipelines.

Previously co-founded Marcellus — bootstrapped the company, grew engineering from 2 to 30, and built an on-demand video transcoding platform serving major media across Asia, the Middle East, and Africa. Served as primary technical advisor to VANS and New Balance at The Marcellus Agency. Earlier: first generation IP Phones at Cisco, adaptive media research at UC Berkeley (Garage Cinema Research → Yahoo! Berkeley Labs), SMIF I/O robotics at Asyst Technologies.

Filed patents on constrained quantum optimization and tensor-train networks for medical imaging. Built Prañāda, a retinal screening platform (91.6% accuracy). Proficient across PyTorch, LangGraph, Docker, Kubernetes, AWS/GCP.