The glasses on, the Pi 5 in its bamboo case in a pocket, and the orange ribbon cable joining the two.
I'm a mechanical engineering student at Nottingham Trent. My family lives in Hong Kong and we're originally from Bangalore. I like taking an idea all the way to something you can hold. I sketch it, model it in CAD and write the code for it, then print, machine and finish it myself. Alongside uni I'm doing the MIT IDSS programme in AI and data science.
Six projects from 2022 to 2026, seen through the lens. Each one has its own stop on the circuit board inside the mini-PC, down the cable.
A-Level D&T project. Python, OpenCV, Raspberry Pi 5, CNC.
GCSE D&T project. Resin printing, leatherwork, C++.
MIT IDSS. K-means, random forest, Jupyter.
Yootun internship, Changsha. Research, sourcing, budgets.
Swift. Your wallpapers in a 3D carousel, for Apple silicon.
EPQ dissertation and the NTU group design project.
Glasses that read the page out loud and tell you what's in front of you.
My client had a family member with a visual impairment, so I designed glasses that help with reading and getting around. The camera takes a photo, the Pi pulls the text out in the right reading order and speaks it, and object detection names what's in view. I built a full working prototype and tested it with my client and their family.
They liked that it was light, thin and simple, and reckoned a mass-produced version could sell for around $400. Their main note was to use a material that feels more premium.
A wallpaper picker I liked on Linux, rebuilt as a native Mac app.
Try it: arrow keys, drag, or click. The label modes are the same three the app has, and they never rename your actual files.
I build websites too. This one is hand-written, no templates. Click a screen to open it and scroll.
The Reel page goes with my real app. The other two are concepts, and all three are still being built.
A watch that nudges you to move when you're studying or working from home.
| Item | Part | Qty |
|---|---|---|
| 1 | Resin case, top merged into the body (version 5, no screws) | 1 |
| 2 | Round screen module | 1 |
| 3 | Charging port | 1 |
| 4 | Leather strap, cut and skived by hand | 1 |
| 5 | Buckle | 1 |
| 6 | Leather strap with holes | 1 |
The face shows the time in cyan with a red second hand. The stopwatch starts and stops with a tap. Snake runs on a round grid, arrow keys or swipe. I coded snake in C++ and the stopwatch in Python, and got help from developers on GitHub and Discord for the face.
What keeps me charged.
Buy a used car and turn it into a VIP-style build, with the car and every part inside ¥30,000.
For three weeks at Yootun, part of Sensteed Hi-Tech, I researched the used car market, negotiated a car purchase, shadowed a modding engineer and helped fit the parts. I wrote two full proposals, a 7th-gen Honda Accord and a 2012 Subaru XV, and both were translated into Chinese. Build the Accord here.
Prices are from my Accord proposal, in RMB. Colours and wheel finishes only change the look, the wrap and wheels cost the same either way.
An 8,000-word dissertation on Mars, and a group report on an earth-sheltered home.
An 8,000+ word dissertation I researched and wrote on my own. It covers life support, radiation, farming, robotics, the psychological load on crews, the cost and space law. My answer was that right now it's close to impossible, and that it can't be rushed.
I was co-project manager and led the sustainability research: greywater recycling, rainwater harvesting, composting toilets, solar PV with battery storage and compost heating. Earth sheltering projected up to a 50% cut in energy use.
Newest first. The full details are on my CV.
Online, alongside uni, with projects throughout.
Three weeks of market research, a car purchase and the APEX builds.
An education NGO.
Maths, Physics, Design and Technology, plus the EPQ.
Two end-to-end projects: clean the data, explore it, build models, then turn the results into recommendations.
Grouped customers from a marketing dataset into segments. I used the elbow method and silhouette scores to pick k = 4, then profiled each group so a business could target them. Full marks.
Predicted whether a US visa application gets certified or denied, using the EasyVisa dataset. I compared three models and picked a tuned random forest.
Forty pins, and most of them are things I can do. Hover one and the same pin lights up on the board.
What I write code in, and what for.
The reading and object detection on the glasses, the machine learning notebooks, the watch stopwatch.
Reel, a native Mac app with AppKit and Core Animation.
Snake, running on the round watch screen.
This site: Three.js for the 3D, canvas for the demos, and the board you're flying over.