Experience
Professional experience in performance engineering, cloud systems, and AI infrastructure.
- Performance Engineer — NetApp · San Jose, CA (Jan 2026 – Present)
- Optimizing performance of AIDE (AI Data Engine), a distributed platform for ingesting, cataloging, and serving enterprise storage data for AI workloads.
- Performance and Capacity Engineer — Meta Platforms, Inc. · Menlo Park, CA (May 2024 – Jan 2026)
- Optimized rack power budgets for multiple AI-inference hardware programs → $100M/year in data-center OpEx savings and a standardized budget-optimization framework across Meta’s AI infrastructure hardware programs.
- Drove cross-functional work to raise the power limit of GPUs used by recommendation services → 7% reduction in GPU capacity required with no increase in power budget ($400k/year benefit).
- Delivered power budget for flash-storage racks providing 7.3 EB of storage to support GenAI training.
- Delivered power budget for flash-storage racks providing 9.9 EB of storage to support MySQL and other services.
- Redesigned the Power Sufficiency Simulator pipeline, increasing input accuracy and reducing Thrift service dependencies by 77%.
- Created a machine-level power-estimation methodology for power-capped machines, improving power-capping efficiency by 11% or reducing overage risk by 0.15%, depending on the scenario.
- Developed a technique to estimate power at the power-device level from host-level metrics, reaching 99%+ accuracy and eliminating 0.9%–6% estimation error across power-device types.
- Improved idle-host power-usage estimation accuracy by up to 31%, strengthening rack-level budget planning.
- Research Intern — Cloud Operations Research (CORE), Microsoft Research · Redmond, WA (May 2022 – Aug 2022)
- Designed Kerveros, a novel Azure admission-control system → 95%+ utilization at 99.9%+ availability in production; deployed in Azure and published at OSDI 2023.
- Research Intern — Gray Systems Lab (GSL), Microsoft · Redmond, WA (May 2021 – Aug 2021)
- Built realistic Azure HDInsight benchmarks by reconstructing Spark workloads from query traces; generated privacy-preserving synthetic datasets.
- Graduate Research & Teaching Assistant — Pennsylvania State University · University Park, PA (Aug 2018 – May 2024)
- Research on cloud-systems evaluation methodology (publications at EuroSys 2024 and 2021); teaching assistant for multiple undergraduate courses.
- Junior Software Engineer — Reve Systems · Dhaka, Bangladesh (Oct 2017 – Jul 2018)
- Prototyped a machine-translation system supporting Bangla and 25 additional languages, plus a Bangla NLP platform (translator, spell-checker).