Hey, I am Risheekkumar

I am a Senior Data Scientist at American Express, where I build recommender systems and sequence models that power personalized experiences at scale.

I am interested in foundation models, LLM systems, and reliable AI evaluation. This is where I share what I learn from building them. Outside of work, you will usually find me playing chess.

Open Source

• fastpragma - An easy-to-use Python API for building PRAGMA-style foundation models from profile and event data, from tokenization and pretraining to embeddings and fine-tuning. PyPI · GitHub

• llm-rsa - Recursive Self-Aggregation for LLMs: sample and synthesize multiple candidate answers into a stronger final response. GitHub

Recent Writing