Nilesh Patil

Head of AI at DreamStreet. Previously Head of Applied Research, Dream11.

About

AI leader building deployable AI systems, agentic workflows, and organizational adoption in regulated and large-scale environments. Currently Head of AI at DreamStreet, building compliance-aware AI architecture for investor and trader workflows across research, brokerage, and advisory domains. Previously led applied AI research at Dream Sports / Dream11, including a Columbia University research center collaboration, cross-geography teams across India and New York, and production ML systems at 250M+ user scale.

Particularly interested in AI harness design, developer productivity, and cooking up novel applications from emerging model capabilities and turning them into reliable workflows.

Currently

Head of AI, DreamStreet — compliance-aware AI architecture for SEBI-regulated Indian investor and trader workflows, spanning research, brokerage, and advisory. Translating domain, product, and regulatory requirements into deployable AI architecture, agentic workflows, and user-facing copilots. Driving AI adoption org-wide through training, rapid prototypes, and redesign of existing workflows around agents and model-assisted operations.

Technical focus

  • LLM-based behavior simulation, persona simulators, and agentic evaluators for personalization.
  • Distributed recommendation, content tagging, and text similarity search at ~100M-entity scale.
  • Feature-store systems supporting 250M+ users.
  • Real-time forecasting (~50k+ forecasts) under strict latency constraints.
  • Deep-learning churn prediction over user lifetime trajectories.
  • Self-hosted SLMs and agent tooling across local, GCP, and AWS environments.
  • Compliance-aware AI harness design — audit-friendly, controllable workflows for regulated domains.

Background

M.S. Data Science, University of Rochester (2017). B.Tech, IIT Roorkee (2013). Previously Staff Data Scientist at the Center for Vaccine Biology (Rochester, NY), building automated ML systems for bio-imaging research including 3D reconstruction from hyper-spectral microscopic scans. Earlier roles in mortality forecasting at AXA Insurance and equipment-failure prediction at AbsolutData.

Currently exploring

Frontier model capabilities as building blocks for reliable products, agentic evaluation infrastructure, and developer tooling that survives contact with regulated production environments.

Get in touch

LinkedIn is the fastest way to reach me — linkedin.com/in/ensembledme. Other profiles below.