Prashant R.

About Prashant R.

I work on large-scale LLM training and customization, including continued pretraining (CPT) and supervised fine-tuning (SFT), focusing on making model behavior changes intentional, measurable, and reliable.I guide the design and iteration of experiments and training pipelines in close collaboration with science teams, shaping improvements to pretraining and fine-tuning workflows from concept through evaluation. My work ensures these changes are well-defined, observable, and correct, emphasizing predictable outcomes and consistent performance. Previously, I built and operated high-throughput ML data platforms for audio, video, and text, processing ~1M hours of data per day to support data quality, governance, and evaluation, enabling teams to train models at production scale.I’m drawn to challenges where experimentation, data, and evaluation converge to reliably guide model behavior at scale.

Experience

1. Software Engineer

   Amazon

   Mar 2022 — Present · Boston, MA, US

   Software Engineer working on large-scale ML systems supporting model training and customization.• Built and operated training and data workflows used to adapt models for specific use cases• Owned system design and implementation across the full lifecycle, from initial architecture to production rollout• Worked closely with applied science, product, and platform teams to translate problem statements into reliable systems• Drove technical decisions around data quality, evaluation, and training readiness

Education

• University of Mumbai

  Bachelor of Engineering (B.E.)

  2008 — 2012

• University of Maryland

  Master's degree

  2017 — 2019