Sayan Ghosh
Graduate Research Assistant
Chapel Hill, North Carolina, United States
Summary
👤
Senior🎓
Top SchoolSayan Ghosh is a PhD student and graduate research assistant in NLP at UNC Chapel Hill with a decade of experience bridging academic research and industry R&D. His work focuses on learning from natural language explanations and interactive learning, with prior projects in abstract language understanding and language grounding. He has interned at Amazon, IBM Almaden, Bloomberg, and completed research and engineering roles at Adobe and IIT Kharagpur, bringing practical product-minded rigor to research problems. Sayan has contributed to high-profile open benchmarks like BIG-bench, adding task-specific validation and tooling that improve robustness for large-scale language model evaluation. Known for precise engineering in research code and a strong theoretical grounding, he combines rigorous experimental practice with interest in making model reasoning and explanations more actionable.
10 years of coding experience
3 years of employment as a software developer
Bachelor of Technology (Hons.), Computer Science and Engineering, CGPA 9.54/10.0, Bachelor of Technology (Hons.), Computer Science and Engineering, CGPA 9.54/10.0 at Indian Institute of Technology, KharagpurDoctor of Philosophy - PhD, Natural Language Processing, Doctor of Philosophy - PhD, Natural Language Processing at University of North Carolina at Chapel Hill
Github Skills (5)
python10
apachebench10
data-analysis9
json9
machine-learning8
Programming languages (6)
Github contributions (5)
google/BIG-bench
Jun 2021 - May 2022
Beyond the Imitation Game collaborative benchmark for measuring and extrapolating the capabilities of language models
Role in this project:
Data Scientist
Data ScientistContributions:27 commits, 6 PRs, 14 comments in 11 months
Contributions summary:Sayan primarily contributed to the `abstract_narrative_understanding` task, focusing on utility functions. Their work involved checking and asserting the consistency of proverb lists used in the task examples, ensuring that the lists are the same, and adding assertions for ASCII characters. They also added a print to note the specific task run for 99 distractors to avoid running it for the other cases.
bertmachine-learningbenchmarkmeasuringbenchmarks
sgdgp/sgdgp.github.io
Jun 2020 - Feb 2024
Github Pages template for academic personal websites, forked from mmistakes/minimal-mistakes
Contributions:46 pushes in 3 years 9 months
templategithub-pages-templatemmistakesmistakesjekyll
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.