Marut Pandya
Research at Osmosis (YC W25)
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolMarut Pandya is an ML/AI researcher and engineer with 8 years of experience building large-scale inference and post-training systems for LLMs, reinforcement learning, and high-performance computing. Based in San Francisco, he has driven production-grade work on TensorRT-LLM, vLLM, Triton and distributed GPU serving (RoCE/InfiniBand, NCCL, SLURM) while contributing to model fine-tuning pipelines including distillation, reward modeling and GRPO/DPO. He joined RunPod early as employee #12 and now researches frontier data and RL infrastructure at Osmosis (YC W25), blending hands-on systems optimization with experimental RL environments. A prolific full‑stack engineer, he led major UI refactors and runtime SBOM features in the 5k+ star ThreatMapper project and maintains multi-language algorithm implementations on GitHub. He pairs deep systems-level performance work (caching, parallel data access, DDP/FSDP/Tensor/ pipeline parallelism) with practical product delivery across cloud-native stacks. An active founder and contributor, he’s equally focused on mechanistic interpretability and inference optimization as the next frontiers of responsible, efficient AI.
8 years of coding experience
3 years of employment as a software developer
Bachelor of Engineering - BE Information Science/Studies, Bachelor of Engineering - BE Information Science/Studies at Visvesvaraya Technological UniversityMaster of Science - MS Computer Science, Master of Science - MS Computer Science at Illinois Institute of Technology
English, Hindi, Gujarati
Github Skills (22)
algorithm10
algorithms10
javascript10
python10
c1110
ui-design10
data-structure10
java10
c1710
javas10
uid10
front-end-development10
data-structures10
react10
vulnerability-management9
Programming languages (14)
Github contributions (5)
deepfence/ThreatMapper
Dec 2021 - Apr 2022
Open Source Cloud Native Application Protection Platform (CNAPP)
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:29 reviews, 56 commits, 21 PRs in 4 months
Contributions summary:Marut primarily focused on enhancing the user interface (UI) of the application, specifically within the registry scan functionality. They refactored and modified the registry table, adding image grouping and incorporating pagination. Further contributions included adding new features such as a runtime BOM table and the ability to filter vulnerability scans. The user demonstrated proficiency in React and related UI component development.
cloudsecuritysecopsvulnerability-managementthreat-analysisobservability
ZoranPandovski/al-go-rithms
Oct 2018 - Oct 2018
:musical_note: Algorithms written in different programming languages - https://zoranpandovski.github.io/al-go-rithms/
Role in this project:Full-stack Developer
Full-stack DeveloperContributions:9 commits, 6 PRs, 2 comments in 1 day
Contributions summary:Marut primarily contributed by implementing algorithm solutions in various programming languages, including Python, Java, and C. The commits demonstrate a focus on data structures and algorithms. The user also included a data structure like Hash Table, and a greedy algorithm implementation within the project.
pythoncomputer-scienceprogramming-languagejavascriptalgorithms-and-data-structures
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.