Skip to content
Samit Mohan

Samit Mohan

I'm an AI Scientist at Beltech AI in Bangalore, building computer-vision systems for city traffic: live CCTV, ≤10ms inference, 0.9+ F1 in production. In my free time I rebuild ML from scratch and write engineering blogs.

GitHubLinkedInEmailCV

From scratch

  1. 01
    The math behind deep learning - gradients and the chain rule, from scratch
  2. 02
    Hello world of neural networks - a digit classifier in pure NumPy
  3. 03
    NumPy + PyTorch from scratch - tensors and autograd, then the same net in PyTorch
  4. 04
    Building PyTorch from scratch - a working framework in ~1300 lines
  5. 05
    Attention residuals - why attention and residual streams work
  6. 06
    The annotated microGPT - a GPT built and trained end to end
  7. 07
    What happens when you press 'submit' on ChatGPT - the full pipeline: data, alignment, inference
  8. 08
    Inference engineering - the KV cache, batching, what makes tokens fast

Projects

mathLM

40M GPT trained from scratch on math. Custom BPE, GQA + RoPE + SwiGLU, sparse MoE, SFT + GRPO reasoning RL. Evaluated on GSM8K.

minitorch

PyTorch in ~1300 lines. Reverse-mode autograd, conv2d, optimizers, LR schedulers. 96%+ on MNIST.

mini-code

Built to figure out what's actually inside an agent loop. Turns out it's 250 lines: tool calls, shell exec, streaming.

tennis-analysis

YOLO + ResNet on match footage. Court detection, ball tracking, shot classification. Tested on Wimbledon clips.

Experience

Beltech AI · AI Scientist

Jun 2025 - Present

City-traffic computer vision on live CCTV: ≤10ms inference, 0.9+ F1 in production, 100+ cameras per server.

Genpact · Data Science Intern

Oct 2023 - Dec 2023

Built an NLP resume screener for data-team hiring and a churn-prediction model (POC), plus dashboards and ETL. Alteryx certified.

MakeMyTrip · Machine Learning Intern

Aug 2022 - Oct 2022

Personalization API with multi-armed bandits and an RLHF reward model for itinerary ranking.

Teaching & Writing

aiengg.dev · AI Cohort TA

Dec 2025 - May 2026

TA'd a 16-week AI engineering cohort for 200+ engineers: RAG, agents, fine-tuning, inference.

TensorTonic · Contributor

Ongoing

300+ implementation and system-design problems; the platform serves 40k+ users.

Education

NHCE, VTU · B.Tech Information Technology

2020 - 2024

CGPA 9.4 / 10 · Founder, Coding Club NHCE.