Parsa Kavehzadeh - Applied Scientist at Dialpad

Parsa Kavehzadeh

Applied Scientist at Dialpad

Old Toronto, Ontario, Canada

Join Prog.AI to see contacts

Summary

👤

Senior

🎓

Top School

Parsa Kavehzadeh is an applied scientist and NLP researcher with eight years of experience building and accelerating large language models and multimodal systems across industry and academia. He led inference-acceleration research at Huawei—publishing Sorted LLaMA at EACL 2024—and delivered nested training, adaptive drafting, and dynamic inference architectures that produced multi-fold speedups on 70B and 38B models. His academic work includes UniChart (EMNLP 2023), a vision-language pretraining approach for chart understanding trained on millions of synthetic and real examples, reflecting strength in both data curation and model engineering. Comfortable shipping production ML on cloud platforms and NPU clusters, he blends deep research with practical deployment (Azure, Kubernetes, DeepSpeed, FSDP). Based in Toronto with an MSc from York University, he combines rigorous evaluation and clever engineering tricks—like optimized sub-model loss weighting and balcony-style layer allocation—to close performance gaps while cutting inference costs.

8 years of coding experience

6 years of employment as a software developer

Amirkabir University of Technology

Master of Science - MS, Computer Science, 8.8/9, Master of Science - MS, Computer Science, 8.8/9 at York University

English, Persian

Github Skills (2)

jekyll6

github-pages4

Programming languages (1)

Python

Github contributions (5)

parsareal/parsareal.github.io

Aug 2020 - Apr 2022

Github Pages template for academic personal websites, forked from mmistakes/minimal-mistakes

Contributions:114 commits, 129 pushes in 1 year 8 months

templategithub-pages-templatemmistakesmistakesjekyll

parsareal/A-defense-method-for-textual-adversarials

Mar 2021 - May 2022

Contributions:3 pushes, 2 branches in 1 year 3 months

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.

Request Free Trial