September 2019 - November 2021
AI & ML Lead
Hike - India
Operating AI/ML leadership role. Computer vision, NLP, recommendation, personalisation and trust-and-safety systems serving more than 100M users. 60+ provisional patent filings.
Hikemoji, the first of the five systems, generated personalized avatars directly from user selfies: facial-feature extraction, a component-matching algorithm, and style-transfer techniques, running on Python, TensorFlow, PyTorch, and OpenCV, with BigQuery for storage and Airflow for orchestration. The other four systems — vernacular NLP, social matchmaking, trust & safety detection, and gaming matchmaking — shipped from the same ML team over the same period.
What I did
- · Led development of five ML systems on a platform with more than 100M users: computer-vision-driven avatars (Hikemoji), vernacular NLP, social matchmaking, trust & safety classifiers, and gaming matchmaking.
- · Delivered Hikemoji, a facial-recognition-driven avatar and social-matching system.
- · Filed 60+ provisional patents and published applied ML research.
FAQ
What did Dipankar Sarkar do at Hike?
AI & ML Lead at Hike (September 2019 - November 2021), leading computer vision, NLP and recommendation work on a platform with more than 100M users, and filing 60+ provisional patents.
What systems did Dipankar build at Hike?
Five production ML systems: computer-vision-driven avatars (Hikemoji), vernacular NLP, social matchmaking, trust & safety classifiers, and gaming matchmaking.
Did the work at Hike result in patents?
60+ provisional patents were filed covering the ML systems built during this period, alongside published applied ML research. These are provisional filings, not granted patents.
Related writing
- Machine Learning at Hike: Five AI Systems
The full write-up: Hikemoji, vernacular NLP, social matchmaking, trust & safety, and gaming matchmaking.