About

Contributed to Vajra, an open-source LLM inference engine, working on quantization & MoE support, experiment infra, CI, and telemetry. Led development of Veeksha, a benchmarking framework for evaluating inference systems with realistic workloads.

Advising companies on end-to-end applied AI: requirements, data, training, eval, and serving.

Studied advanced batching for LLM inference during my MSc.

ML systems for asset & credit risk. Productionized work with measured impact in the tens of millions.

~2 years across two projects. Explainable AI research (evaluation of attention-based explanations) under Dr. Alvin Jia and Prof. Albert Bifet. Data analysis and modeling for COPEDI-Cat, a Catalan COVID-19 paediatric response network (~150 collaborating paediatricians).

First cohort of Spain's data science program. Thesis and internships centered on ML interpretability.