TM
Thilak Mohan
Workspace / Thilak Mohan / About
Mode
Accent
Font

Thilak Mohan

Talk to users first, build second. See why this is my motto here.

Status
Working at Promptli AI
Location
San Francisco, CA
Role
Applied AI Engineer
Stack
PythonPyTorchTypeScriptReactLangChainAWS

Origin — a tool for my mum

How I learned that the shortest path to product-market fit is being one call away from a real user.

AI-native and Montessori pedagogy-aligned SaaS to copilot teachers.

Read more →
4
Schools · 2 states
100+
Active teachers
6,000+
Notes / month
1hr+
Avg session time

Selected work

Filtered: shipped. Hover for a preview, click for details.

Project
Year
Role
Context

How I work

Things I've learned from experience shipping to prod.

Lab — GPT-2, live

A tiny demo of the 128M-param transformer I built from scratch. Prompt it below.

🔬 These models were trained on 4 A100 GPUs for 2 days, so keep your expectations… modest. Built purely for educational purposes — I get to dig deep into PyTorch implementations and learn firsthand. Functionally useless? Maybe. Incredibly valuable for learning? Absolutely.
🧠
thilak/gpt2-small
128M params · 10B tokens · 4×A100
live
🧠
model.eval() · 128.2M params loaded · awaiting prompt
gpt2-small · just now

Resume