Piotr Dabkowski
About
Detail
Co-Founder
United States
Résumé
Jobs
verified_user
0% verified
Co-FounderElevenLabsJan 2022 - Current (4 years 9 months)- Software EngineerGoogleMar 2018 - Jan 2022 (3 years 11 months)
- Software EngineerTessianNov 2017 - Feb 2018 (4 months)
- Software Engineering InternGoogleJul 2016 - Sep 2016 (3 months)
- Software Engineering InternOperaJul 2015 - Jun 2016 (1 year)Developed automated user data analysis pipeline. Implemented a system automatically collecting the data from multiple sources and later predicting user conversion rates and their lifetime values using ML. I was working full-time in the Oslo office during summer and part-time from Oxford during term time.
Education
verified_user
0% verified
- Master of Philosophy - MPhil, Advanced Computer ScienceUniversity of CambridgeJan 2016 - Dec 2017 (2 years)MPhil Thesis under supervision of Prof. Yarin Gal: ''Real-Time Image Saliency For Black Box Classifiers'' (published at NeurIPS 2017) https://arxiv.org/abs/1705.07857
- Bachelor of Arts - BA, EngineeringUniversity of OxfordJan 2013 - Dec 2016 (4 years)
Projects (professional or personal)
verified_user
0% verified
- JJs2Py - JavaScript implementation in pure PythonSep 2014 - Current (12 years 1 month)A complete implementation of JavaScript in pure Python. Over 2M installs, and ~2k stars on GitHub.
Publications
verified_user
0% verified
- NReal-time image saliency for black box classifiers (NIPS 2017)NIPSMay 2017In this work we develop a fast saliency detection method that can be applied to any differentiable image classifier. We train a masking model to manipulate the scores of the classifier by masking salient parts of the input image. Our model generalises well to unseen images and requires a single forward pass to perform saliency detection, therefore suitable for use in real-time systems. We test our approach on CIFAR-10 and ImageNet datasets and show that the produced saliency maps are easily interpretable, sharp, and free of artifacts. We suggest a new metric for saliency and test our method on the ImageNet object localisation task. We achieve results outperforming other weakly supervised methods.