I build ML systems that move metrics at scale — ranking, personalization, and pricing for 1M+ monthly users at Cars24, and India's first SEBI-compliant agentic trading platform at KotiLabs. 4+ years across marketplaces, fintech, and AI products. Open to FAANG and high-growth startups.
This site has the full case studies. The PDF is a recruiter-friendly, one-page summary.
A full 0→1 system that takes a user's voice command ("Buy Tata Steel if it breaks VWAP") through intent classification, LLM planning, compliance validation, strategy backtesting, and live broker execution — with zero hallucination risk at any step. Built to meet SEBI's April 2026 algorithmic trading circular.
Rebuilt the recommendation engine using a Two-Tower retrieval model with HNSW ANN index, replacing a 25-cohort system with individual user-level rankings at 1M+ monthly sessions. Solved cold-start for 35% of zero-click users via implicit search/filter signals — first time the platform achieved personalisation for this segment.
Supervised regression model using vehicle fingerprint, demand/supply signals, market science, and LLM-extracted inspection quality scores to predict optimal listing price.
My M.Tech thesis on domain-specific transformer adaptation for legal NLP — the same principle I apply in production today.
Demonstrated that domain-specific pre-training (Legal-BERT) substantially outperforms general BERT on Indian legal case judgment prediction tasks. Established a benchmark for domain-adapted transformer models in legal NLP, validating the intrinsic dimensionality hypothesis: domain-specific adaptations occupy a low-rank subspace of the weight space — the same principle behind LoRA fine-tuning, which I apply at KotiLabs for intent classification.
I'm actively looking for Applied Scientist, AI Engineer, MLE, and Senior DS roles — FAANG and high-growth startups both. I respond to well-matched opportunities within 24 hours.