Catalin Voss
Co-Founder & CTO at Smile Identity
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolCatalin Voss is a serial founder and technical leader with 12 years of experience building AI-driven products and startups, currently serving as Co-Founder & CTO of Ello in San Francisco. He blends deep research-level expertise from Stanford (PhD in AI, on leave) with hands-on engineering, having led and sold ventures like DukaConnect and Sension and contributed production-grade code to flagship open-source ML projects such as Hugging Face Transformers and Mozilla DeepSpeech. His work spans computer vision, speech, and ML infrastructure—practical systems that run on-device and at scale—and he’s advised startups and VCs on productizing research. Notably, he spearheaded the Autism Glass Project at Stanford, demonstrating a track record of turning academic research into real-world assistive technology.
12 years of coding experience
8 years of employment as a software developer
High School Abitur, High School Abitur at Leonardo da Vinci Gymnasium, Neckargemünd, GermanyPhD (on leave) Artificial Intelligence, PhD (on leave) Artificial Intelligence at Stanford University
English, German, Romanian, Spanish
Stackoverflow
91reputation
1kreached
1answer
3questions
Github Skills (34)
transformers10
pytorch10
language-model10
postgresql10
python10
data-science10
machine-learning10
mozilla-deepspeech10
tensorflow10
bert10
speech-recognition10
gpt10
nlp10
pre-trained-model9
multithreading9
Programming languages (9)
Github contributions (5)
mozilla/DeepSpeech
Nov 2020 - Apr 2021
DeepSpeech is an open source embedded (offline, on-device) speech-to-text engine which can run in real time on devices ranging from a Raspberry Pi 4 to high power GPU servers.
Role in this project:
ML Engineer
ML EngineerContributions:24 reviews, 88 commits, 13 PRs in 5 months
Contributions summary:Catalin primarily contributed to the DeepSpeech speech-to-text engine by making changes to the training scripts and utility functions. Their work involved modifications to the inference graph, incorporating remote I/O capabilities for file access, and optimizing audio processing components. They also addressed imports and dependencies and refactored some aspects of the code.
speech-recognitiontensorflowreal-timeraspberry-pion-device
snorkel-team/snorkel
Jul 2018 - Feb 2019
A system for quickly generating training data with weak supervision
Role in this project:Back-end Developer & Data Scientist
Back-end Developer & Data ScientistContributions:26 commits, 10 PRs, 18 comments in 7 months
Contributions summary:Catalin primarily contributed to the back-end of the Snorkel system by optimizing database interactions and improving the efficiency of data processing. Their work involved adding indexes to foreign keys to speed up cascade deletes and refactoring the progress bar functionality within the UDF runner to support multi-threading and notebook compatibility. Additionally, the user made modifications related to scoring metrics within the learning utilities, suggesting an active role in the machine learning aspects of the project.
weak-supervisionpythondata-sciencemachine-learninglabeling
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.