Teach models to read people
My research is video models that judge a person's state, tested so the numbers hold outside the lab. One IEEE paper and two awards so far.
The researchHi, I am Abhijith. Learning to build machines that notice people.
Systems engineer in London, working on perception and control.
So far that means video models that try to tell when a driver is drowsy, and the estimation and control that lets a machine act on what it sees. First-class BEng from Warwick, ACSE MSc at Imperial, one IEEE paper, three apps other people use every day.
Systems engineerAbhijith PradeepLondon
- PyTorch
- OpenCV
- Python
- C++
- MATLAB
- TypeScript
- React
- Flutter
- Supabase
- Cloudflare
- ESP32
- numba
What I do
Make machines act on it
Kalman filters and PID, LQR and MPC on real hardware: a rotary pendulum, a landing drone, and a ball-on-plate rig built to compare controllers fairly.
The balancerShip things people use
A maths tutor, a shop's staff system and a test trainer, all live and used every day by people who are not developers.
The live appsMeasure before I believe
I build the benchmark or the test harness first, so a result can be trusted and a failure can be named. It is the habit that got a chess engine to 7th on a ladder of 465.
The chess recordThree things to look at first
Live in your browser
The balancer's control loop
This is the loop from my ball-on-plate project. A camera watches the ball, a Kalman filter works out where it really is from the noisy readings, and a PID controller tilts the plate to bring it to the target.
- the ball
- the target. Click the plate to move it
- what the camera reports, with noise
- what the filter believes
Every few seconds the ball gets a kick, so you can watch it recover. The plate's shadow shows which way it is tilted.
- controller
- PID
- tilt x
- 0.0°
- tilt y
- 0.0°
- error
- 0 mm
- estimator
- Kalman
88%Dissertation mark, IMechE Best Project
3Apps other people use every day
11Finished projects, and 6 more on the way
47thOf 465 on the Chessathon ladder, best rank 7th
Experience
Open a role for what it involved.
Mar 2025 to present Coding Analyst, Outlier AI
I evaluate how language models handle coding and reasoning tasks, over 1,000 outputs so far, and write prompts designed to make them fail, over 100 of those. The interesting part is the edge cases.
Aug 2025 to Feb 2026 Student Trader, Maverick Student Managed Fund
Built backtesting in Python to test hedging strategies and risk measures (VaR, drawdown, Sharpe) across more than 10 positions and five years of data. Tested five systematic strategies; the best cut simulated slippage by 10 to 15 percent.
Jul to Sep 2025 Instrumentation and Control Intern, SBM Offshore
Read more than 30 piping and instrumentation diagrams, found where the documents disagreed with API and IEC standards, and kept over 100 I/O list entries and instrument indexes current.
Oct 2023 to Apr 2024 App Development Intern, Applied Biomedical Signal Processing and Intelligence Lab, Warwick
Helped build a Python and C++ algorithm that detects eight kinds of cardiac arrhythmia from ECG signals, and designed the screen that shows its predictions to the person using it.
Education
MSc Applied Computational Science and Engineering (ACSE)
Imperial College London, 2026 to 2027.
BEng Systems Engineering, First-Class Honours
University of Warwick, 2023 to 2026. Dissertation at 88 percent, IMechE Best Project Certificate 2026.
Leadership
- Elevayte Co, founder. A non-profit in Bangalore. Ran a fundraiser with two sponsors and over 250 participants.
- Warwick BORING Society. Launch-system optimisation, US shipment logistics, volunteer coordination.
- Warwick Sustainable Finance Society. Led the strategic relaunch and sponsor engagement.
Skills
Hover or tap a skill to see where I used it.
Perception and machine learning
Control and systems
Software and product
Working with models
Hover or tap a skill to see where I used it.
Interested in
working with me?
Open to research roles and collaborations in perception, control and applied machine learning. Based in London.