Bjarke Felbo
Cofounder & CTO at Recognition Labs
Copenhagen, Capital Region of Denmark
Summary
🤩
Rockstar🎓
Top SchoolBjarke Felbo is an AI researcher-turned-founder and CTO with 11 years of experience building consumer-facing products and production-grade ML systems from research prototypes to 10M+ installs. He co-founded Rune and RuneMatch, designing infrastructure, real-time voice/chat, embedded HTML5 games, and AI-driven moderation that together created cohesive social experiences, and now leads a stealth AI venture out of Copenhagen. A Marvin Minsky Fellow from MIT, he open-sourced DeepMoji—a widely cited model for sentiment and emotion analysis—which now informs large-scale NLP work and is used to serve hundreds of millions of daily signals. Comfortable bridging deep learning research and engineering execution, he’s equally at home finetuning models and shipping scalable client-facing features. Notably, he pairs academic rigor (MIT, YC, Forbes 30 Under 30) with hands-on product hiring and team building that scales culture as well as code.
11 years of coding experience
12 years of employment as a software developer
Master’s Degree MIT Media Lab, Master’s Degree MIT Media Lab at Massachusetts Institute of TechnologyWinter Batch 2019 with Rune, Winter Batch 2019 with Rune at Y CombinatorMBA Classes (Exchange Student), MBA Classes (Exchange Student) at MIT Sloan School of ManagementTechnical University of Denmark
English, Norwegian, Danish
Github Skills (11)
neural-network10
sentiment-analysis10
keras10
machine-learning10
deep-learning10
tensorflow10
python10
natural-language-processing10
fine-tuning9
text-classification9
nlp9
Programming languages (18)
Github contributions (5)
bfelbo/DeepMoji
Jul 2017 - Sep 2018
State-of-the-art deep learning model for analyzing sentiment, emotion, sarcasm etc.
Role in this project:
ML Engineer
ML EngineerContributions:28 commits, 14 PRs, 26 pushes in 1 year 1 month
Contributions summary:Bjarke primarily contributed to the deep learning model's finetuning and evaluation processes. They fixed typos, corrected architectural method names, and made the emoji mapping clearer within the examples. Their work involved modifying the finetuning functions and tests, and also improving the model's ability to handle edge cases, particularly addressing issues with small attention weights to prevent NaNs.
pytorchnlpanalyzingpythonart
rune/rune
Sep 2021 - Jan 2023
Build multiplayer web games played by millions! 👾 You focus on your game, Rune handles everything else, incl. backend and networking 🚀 Make your game using JavaScript / TypeScript and any frontend web framework such as React, Svelte, Vue, Threejs, PixiJS, Phaser. You can even use plain HTML and CSS. Perfect for web devs interested in gamedev.
Contributions:11 releases, 201 reviews, 111 commits in 1 year 4 months
html5-gamegametinyjavascriptsdk
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.