How an M.Sc. in Mathematics became a tech consultant in the age of AI — and why that combination matters.
BITS Pilani. M.Sc. Mathematics. Five years of abstract algebra, real analysis, topology, and probability theory. At the time, I wondered if I'd ever use it.
Then I started working with machine learning and everything clicked. Linear regression? Least squares optimisation. Neural networks? Chain rule and gradient descent. CNNs? Linear algebra on 3D tensors. Andrew Ng's courses weren't just tutorials — they were a translation layer between mathematical theory and computational practice.
One of my earliest projects: indoor navigation using BLE beacons — Changi Airport style. The algorithm? Kalman filtering — predicting position from noisy sensor data with under 2-metre accuracy across 50,000 ft².
Pure applied mathematics: probability distributions, Bayesian updates, state estimation. Same math used by self-driving cars. Same math used by LLMs for next-token prediction. Different scales, same foundations.
Most AI consultants learned prompt engineering. I learned gradient descent.
When a startup says "we need AI," what they usually mean is: the data pipeline is broken, the architecture won't scale, or they don't know what "good" looks like.
A math background means I can diagnose why models fail (bias-variance trade-off, loss landscapes), design data architectures that scale (distributed systems is probability theory + graph theory), and cut through the hype (transformer architecture is matrix multiplication with attention weights — not magic).
I built Ekam — a programming language compiling to Zig — because the mathematician in me wanted to understand computation at every level. When you hire me, you're getting someone who understands the principles underneath, not just the latest framework.
If you're building something that needs deep thinking — not just deep learning — let's talk.