About me
I'm a research engineer working in medtech at IIT Madras' Healthcare Technology Innovation Centre — the kind of engineering where a millimetre is the tolerance and a few milliseconds is the whole budget, because the software ends at a surgeon's hand. It taught me the habit I care most about: writing code that has to be right, and then proving it is.
I did not arrive here in a straight line. I came second in the state in the NMMS national examination, finished my Computer Science degree with a 9+ CGPA and the Best Outgoing Student award, and spent the years in between shipping applied machine learning — spam classifiers, price regressors, housing pipelines — always taken past the notebook and served behind a real API, because a model nobody can call is a model nobody uses.
Somewhere in there I also learned to publish. Two papers now, one with IEEE, one with a DOI — and a runner-up placing at an international conference for presenting the second. Writing research turned out to be the same discipline as writing tests: say precisely what you did, and show why it holds.
These days I mentor with Girls Leading Tech, which is easily the most useful hour of my week. I remember exactly how opaque this field looked from the outside, and how much of the barrier was information nobody thought to hand over. I'm currently open to software and machine learning engineering roles.
-
2
Papers published
-
16
Certifications
-
9+
CGPA / 10
-
2nd
State rank, NMMS
What I do
-
GPU & real-time systems
CUDA kernels and Holoscan operator graphs that close a surgical loop inside a single frame, without ever round-tripping to the host.
-
Machine learning research
Neural signed distance fields, probabilistic point-set registration, and the validation work that turns a result into something clinicians can rely on.
-
Full-stack engineering
Next.js and FastAPI products with real data behind them — retrieval over pgvector, auth, and models served through APIs people actually call.
-
Numerical validation
Test suites for code where being approximately right is not good enough — including byte-identical determinism checks across runs and machines.
Published research
-
IEEEAthercare: Unified AI System for Health Intelligence in Real Time
6th International Conference on Smart Electronics and Communication (ICOSEC 2025), Kongunadu College of Engineering & Technology, Trichy
Quality healthcare access is a worldwide problem — disjointed systems, slow diagnostics, and dense medical documentation all sit between people and care. Athercare sets out a unified digital health ecosystem that simplifies and improves how that care is delivered.
-
IJATEMPredictive Maintenance of Industrial Equipment Using Temporal Graph Neural Networks
International Journal of Advanced Trends in Engineering and Management — ISSN 2583-7052
Industrial equipment fails on its own schedule, and fixed maintenance intervals either arrive too late or waste working parts. This paper models the plant as a temporal graph, so failure prediction can draw on how equipment relationships evolve rather than on each machine's history in isolation.
View DOI
