Data & AI Engineer. Co-Founder at Autostrata.ai. Royal Holloway MSc, Merit. Published researcher. Currently shipping ML and full-stack products for paid clients.
Got a question?in the corner ↘
I started in Delhi, learning data and machine learning between IIT Madras and Jamia. I moved to London in 2023 for an MSc at Royal Holloway, and stayed to build.
Today I'm a Data & AI Engineer and Co-Founder of Autostrata.ai, a UK studio that ships marketplaces, payment infrastructure, and AI agents for founders and SMBs. I split my time between client delivery, original research (a published preprint on football sentiment analysis), and personal data stories that nobody asked for but I couldn't stop thinking about.
CurrentlyBuilding agentic AI patterns for clients via Autostrata, and writing about what I learn — most recently, the AI assistant on this site.
MSc Data Science & Analytics
Royal Holloway, University of London · Merit · 2024
Visa status
UK Graduate Visa, valid through January 2027. Currently building from Egham, Surrey.
Real projects with real clients, paying real money. Every metric below comes from a live deployment, not a demo.
Self · taiyab.autostrata.ai
An agentic AI you can actually try, right here. Same architectural pattern as the HEART AI I built for a paid client.
Paid client · via Autostrata
SKU-level proof-of-origin for handwoven textiles — so ethical fashion brands can prove their claims before EU law forces them to.
Paid client · via Autostrata
Face-verified, GPS-geofenced attendance for a 45-employee operation.
Paid client · via Autostrata
Voice-activated SOS app with on-device wake-word and speaker verification.
Paid client · via Autostrata
A custom Sharetribe Flex marketplace for designer outfit rentals.
MSc Dissertation · Royal Holloway, University of London
The model with the worst benchmark score succeeded where the best one failed.
Earlier client work
Paid client · via Autostrata
The production marketing site and secure waitlist backend for a pre-launch UK tax-automation fintech — plus a full audit of its AI-generated codebase.
Gurucool XYZ Pvt Ltd
Hybrid recommendation engine that lifted engagement 50% and cut churn 40%.
Published research, personal data stories, and tools I built to solve problems I was experiencing. The work that doesn't fit on a CV bullet point.
“The model with the worst benchmark score succeeded where the best benchmark model failed. Because 79% of football tweets are actually neutral — and the binary classifier forced all of them into ‘positive’ or ‘negative’, burying the real signal.”
Personal research
Is the modern NBA really just more threes? The numbers say it's a structural shift.
Open-source tool
I built a tool to solve the problem I'm still living through.
Personal research · finance module spin-out
When you're not clear on the question, more metrics just add confusion.
BSc Final Project · Jamia Millia Islamia
Computer vision classifier on MRI imagery — 84% accuracy, F1 = 0.8.
No padding — these are the tools I've actually shipped with in the last twelve months. Daily-driver tools at the top of each group.
I'm available for full-time data, ML, or AI engineering roles, and for freelance / contract work via Autostrata. Always happy to hear about research opportunities too.
© 2026 Mohammad Taiyab Khan. Built with Next.js + a lot of coffee, in Egham.
v1.0 · taiyab.autostrata.ai