Sparsh Jain

At American Express, I build enterprise AI systems for credit risk, fraud, marketing, and financial document automation. I architected a multi-agent framework for autonomous feature discovery that reduced manual feature engineering by 70–80% and generated 200+ interpretable feature candidates.

I advance foundation models for tabular and time-series prediction, improving model performance by 10–15% across enterprise portfolios. I’ve also built multimodal LLM and RAG pipelines that reduced manual financial-statement processing by 60%+, alongside an internal LLM code assistant that automated roughly 40% of engineering queries.

I bring research rigor to production AI through fairness-aware learning, differential privacy, explainability, governance, and scalable ML platforms. My ICML 2021 research established statistical guarantees for fairness-aware learning, and my string-matching work has generated $5M+ in annual savings per use case.