I'm a machine learning software engineer with experience in building data pipelines, training open-source models, and deploying them in production. I grauduated with a Master's in AI from the University of Edinburgh (batch 2022), where I explored the potential of self-supervised learning techniques in learning from erroneous transcriptions. These techniques helped in training new ASR models with improved performance across various language and acoustic model configurations.
I've mostly worked in speech and conversational AI, produced some research work [Google Scholar], and now I'm diverging to learning/solving problems like matchmaking and recommendations, LLMs, and ML lifecycle management toolings.
🍃 Matchmaking simulation engine here
Motivation: - Frustrated by League of Legend's matchmaking ⚔️, I decided to start working on my own game simulation to understand matchmaking and ranking algorithms, and eventually build a working software.
- Short term:
- Built a client-server using FastAPI REST and websockets.
- Working on dockerizing the application, kubernetes for orchestration and redpanda for live streaming.
- Long term:
- For matchmaking use: A Bayesian Approximation Method for Online Ranking by Weng and Lin.
- have plugins for different matchmaking algorithms: TrueSkill, Elo, (find more) (try to match league's MMR)
🌱 Open Source Contrutions:
- I'm helping with dashboard backend, resource organization, and platform optimization improving user experience @FeatureForm, the Virtual Feature Store. All with the kind mentorship from their wonderful team.
- At HuggingFace, I'm working with Diffusers and contributing to their vision models.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent