Headline

ML Systems Engineer | Post-Training, Evals, VLMs | Amazon GenAI

About

I build machine learning systems for frontier-model workflows i.e. post-training, evaluation, reward design, VLMs, data curation, and inference. My recent work spans production LLM/VLM systems, including post-training, designing reward functions, building eval frameworks, and creating large-scale data pipelines for foundation-model training. I am especially interested in post-training, RL, VLMs, evals, synthetic data, data curation, and applied ML infrastructure. I also built Veridis, a legal AI workspace, as a way to apply these ideas to real messy knowledge work: long documents, evidence retrieval, drafting workflows, evaluation, and human-in-the-loop review. Looking for frontier AI engineering roles where the work is close to models, data, evals, and product behavior.

Location

Greater Bengaluru Area

AI stack

anthropic apis

antigravity

claude code

claude

codex

cursor

Work experience

Software Development Engineer II at Amazon, Aug 2022 - Present · 4 yrs 1 mo, Bengaluru, Built production AI and platform systems across GenAI, model workflows, evaluation, data pipelines, inference, and large-scale consumer product infrastructure. - Worked on LLM/VLM workflows spanning post-training, reward functions, evals, data curation, and inference optimization. - Built foundation-model data pipelines with filtering, deduplication, and PII scrubbing. - Built eval and experimentation loops to measure performance of the developed and served model

Software Engineer at Paytm, Jul 2021 - Jul 2022 · 1 yr 1 mo, India, Built data and search infrastructure for merchant intelligence over large-scale payment streams. - Built Spark feature pipelines over merchant transaction streams - Built Elasticsearch-based similarity search for merchant segmentation using behavioral embeddings and ranking signals. - Reduced ML pipeline compute cost by 90% through Spark repartitioning, reusable feature snapshots, and parallelized backfills.

Software Associate at JPMorgan Chase & Co., Jan 2019 - Jul 2021 · 2 yrs 7 mos, Built pricing, valuation, and cost-optimization systems for financial derivatives workflows. - Developed a collateral pricing and valuation platform across collateral positions, haircuts, margin rules, and market-data inputs. - Built workload and cost models for valuation runs, saving $400K+ annually through better compute scheduling.