At Straive, I led client-side product discovery for PGIM, capturing requirements and aligning technical scope with alternative investment workflows. I built an AI-assisted entity resolution platform that unified investor entities across three siloed workstreams, with an agentic system for ambiguous matches that achieved a 0.93 F1 score.
I also built an agentic chatbot that combines internal CRM records with public market data to match investor interests to financial vehicles. Its feedback loop learns from sales representatives’ approvals and rejections to refine prompt taxonomies and scoring.
At Handshake AI Solutions, LLC, I collaborated with a leading AI lab firm on LLM fine-tuning through RLHF and SFT. I created labeled ML-agent trajectories and adversarial behaviors to support evaluation and test alignment classifiers.
Earlier, I developed a multi-agent reinforcement learning system for Monster Gridiron using Unity ML-Agents, and an intelligent file organization workflow using GPT-4o and LangGraph. At Antriksh Labs Pvt. Ltd., I created a sensor failure prediction system that identified 85% of anomalies.
