István Fehérvári
Senior Staff Machine Learning Engineer at Juniper Square
North Vancouver, British Columbia, Canada
Summary
🤩
Rockstar🎓
Top SchoolIstván Fehérvári is an AI engineer and scientist leader who co-founded and serves as CTO of a stealth generative AI startup focused on next-level personalization while also acting as Director of Data & ML at BenchSci and Chief Scientist at Ingram Technologies, based in North Vancouver. He holds a PhD in Computer Science (distinction) and brings about 10 years of industry experience building production ML systems, including leading ML teams at Loblaw and earning two patents at Amazon for ML in real-time video encoding. A hands-on contributor to open-source deep learning projects like Apache MXNet and Gluon-CV—where he added operators and fixed tricky bugs—he also spans full-stack work demonstrated by integrating a Hearthstone deck importer. Known for scaling teams and mentoring technical hiring (participating in ~100 hiring loops), he combines academic depth with pragmatic, product-driven engineering.
10 years of coding experience
16 years of employment as a software developer
MSc, Integrated Engineering, MSc, Integrated Engineering at Budapest University of Technology and EconomicsMathematics, Mathematics at Teleki Blanka GimnáziumPhD, Computer Science, Distinction, PhD, Computer Science, Distinction at Universität KlagenfurtComplex Systems, A, Complex Systems, A at New England Complex Systems Institute
English, German, Hungarian
Github Skills (36)
ui-d10
ios10
operation10
python10
tensorrt10
mxnet10
machine-learning10
gluon10
ui-design10
numpy10
deep-learning10
api10
tensorflow10
front-end-development10
computer-vision10
Programming languages (8)
Github contributions (5)
HearthSim/HSTracker
Aug 2016 - Feb 2019
A deck tracker and deck manager for Hearthstone on macOS
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:16 releases, 276 commits, 48 PRs in 2 years 6 months
Contributions summary:István implemented and integrated a deck importer feature from hearthstonetopdeck.com, adding functionality for users to import decks directly into the Hearthstone tracker application. This involved creating new UI elements for the importer, handling potential import errors with localized messages, and integrating the importer with the application's database for managing imported cards. The user also made various code adjustments to the application's database, importer and UI components.
decktrackertilesethearthstonedeck-manager
apache/mxnet
May 2018 - Mar 2019
Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
Role in this project:ML Engineer
ML EngineerContributions:9 commits, 13 PRs, 56 comments in 10 months
Contributions summary:István primarily focused on improving the `mxnet` deep learning framework, specifically addressing bugs in existing modules and implementing new functionalities. Their work included fixing a division-by-zero bug in the `DistanceWeightedSampling` example and enhancing the `ColorNormalizeAug` image augmentation function. Additionally, the user added the `diag()` operator to the framework and generalized the `broadcast_like` operator, contributing to the expansion and refinement of the library's capabilities.
pythonschedulerdataflowmutationdata-science
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.