§ 0
— mathematics · research · tools
Adithya
Kumar
I study mathematical problems and build tools for research. The projects here move between theory, simulation, computer vision, and hardware, following the practical questions that connect them.
Work
Descrai v3: Lab Intelligence
50% of published research is irreproducible because the raw process data is never captured. Descrai v3 is a VLA pipeline that takes experimental notes and verifies protocol adherence in real time. To get it into labs, we built smart safety glasses — giving the system eyes and ears at the bench.
Lasserre Hierarchies for MAX-CUT
Dissertation: derived and compared Lasserre SDP relaxation hierarchies across multiple MAX-CUT formulations, examining how tightly successive levels bound the optimum and where formulations diverge in quality and cost.
eGates Queue Simulation
Modelled immigration queuing dynamics at an international airport using queuing theory. Treating passenger flow as a continuous fluid rather than discrete events cut computational cost without losing the dynamics that determined throughput.
Only undergraduate entrant among 100+ postgraduate research teams
About
A collection of questions, experiments, and things made along the way.
I'm studying Mathematics at the University of Edinburgh. My dissertation looks at the Lasserre hierarchy: semidefinite-programming relaxations for MAX-CUT and related combinatorial optimisation problems.
I'm drawn to work where an abstract model has to meet a real setting: simulations of airport queues, learning methods for graph problems, and computer-vision systems for physical spaces.
More recently, that has meant building Descrai: software and hardware intended to help labs capture the details of their experimental work. I tend to learn the next tool when the work calls for it.
Current threads
Mathematical optimisation
Semidefinite programming, combinatorics, and computational questions.
Research tools
Systems that make experimental work easier to record and review.
§ 3