Tanay Mehta

ML Engineer at CNTR

My work spans agentic systems, language model training, and inference performance.

Based in
Frankfurt am Main, Germany
Kaggle milestone
Notebooks Grandmaster
Tanay Mehta in front of Tower Bridge

About me

I taught myself to program and got into machine learning through Andrew Ng’s courses. I started competing and sharing notebooks on Kaggle in 2020, during the lockdown in India.

I’m interested in what happens underneath the API, from how backpropagation computes gradients to where a language model spends time on the GPU. I’ve explored those questions by building an automatic differentiation library in NumPy, profiling GPT-2, and writing a fused CUDA kernel. I share what I learn through open-source code and technical writing.

Experience

CNTR

ML Engineer

Joined as CNTR’s first technical hire to build ML systems that help people and AI work together.

Aleph Alpha

AI Engineer · Full-time

  • Helped architect and scale enterprise agentic systems on the core Agentic Chat team.
  • Contributed to the Learning team’s fine-tuning API for distributed LLM pre-training and on-demand fine-tuning.
  • Worked on search for agentic systems, available as an MCP tool.

University of Bath

Graduate Teaching Assistant · Part-time

Helped undergraduates with natural language processing concepts and coursework for CM30320.

Education

MSc in Data Science

University of Bath · Distinction

For my dissertation, I trained language models at different scales to study scaling laws and look for emergent abilities such as mathematical reasoning. Supervised by Dr Nello Cristianini.

BTech in Computer Science

JECRC University · First Division

Minor in Mathematics & Engineering. Final grade 8.76 / 10 (87.6%).

Open Source Projects

GitHub