Abhishek Vermma

At NoBroker, I lead AI and machine learning initiatives including RAG-based Chat with Data and multi-agent workflows. My Call Report Copilot reduced manual reporting effort by 50%, delivering approximately $5M in quarterly savings.

I build and fine-tune production ML and GenAI systems using Python, RAG, Azure OpenAI, Azure AI Search, PEFT, and modern MLOps practices. Through prompt optimization, embedding tuning, and retrieval improvements, I achieved 286/300 accuracy in human evaluation.

Previously, at Bridge Solutions Group and Yantra Commerce, I analyzed large-scale supply-chain data with SQL and Python and created Tableau and Power BI dashboards for orders, inventory, fulfillment SLAs, and operational KPIs.

I'm also interested in deep reinforcement learning research, developing transfer-learning and contextual-bandit approaches with PyTorch and OpenAI Gym that improved generalization, training time, and Atari benchmark performance.