About

About me#

I work on applied AI for healthcare. Two lines of work occupy most of my time. The first is the perception stack for medical devices being built in collaboration with clinical partners — an AI-enabled vaginal speculum for cervical screening (Visispec), a joint effort between Odyssey Therapeia and Punarbhava under the PxO partnership. The second is clinical-psychology research with a focus on adolescent mental health: literature reviews, dataset preparation, and contributions to AI-assisted therapeutic models. The two lines share a discipline — measurement on real populations under conditions a teaching hospital does not always reproduce — and they share a lab.

I joined Nautilus Lab — Odyssey Therapeia's research unit — to build the parts of these systems that have to work outside the lab. That means computer vision pipelines that stay reliable across imaging conditions, language-model interfaces that physicians actually use, and the unglamorous infrastructure (data pipelines, evaluation harnesses, monitoring) that keeps deployed models honest.

Current focus#

Computer vision in healthcare. Robust segmentation and classification on small, heterogeneous medical datasets. Domain shift between training data and clinical-site data is the hard problem; everything else follows from getting it right. The Visispec work — accepted at the ESGO 2026 Congress and published in the International Journal of Gynecological Cancer — is the central artefact in this line.

LLMs for translational medicine. Using large language models as interfaces to structured medical knowledge. The interesting work is not in the model — it is in the constraints, the retrieval, and the ground-truth signal that prevents plausible-sounding wrong answers from reaching a clinician. I wrote a 2025 review of AI-driven enhancements in non-invasive prenatal testing as part of this line, and a 2024 piece on cloud-driven interoperability in healthcare with co-authors at Dronacharya.

Multi-agent systems for robotics. The REPAM framework, with Rachit Sehgal and the team, addresses a hybrid multimodal-model operations problem in robotics. It appeared in ICT for Intelligent Systems (Springer Nature Singapore, 2026).

Medical devices and embedded AI. What changes when the model has to run on a device under real power and latency budgets, and what changes again when that device is in a clinical environment. Most ML papers do not address this; most of my recent engineering effort is about closing that gap.

How I work#

I prefer carefully held-out evaluations to leaderboard chasing. I write most research code in PyTorch, with a side stack in JAX where second-order optimisation matters. For clinical projects I treat the evaluation harness as part of the deliverable: a model is only useful if its calibration on new sites is well-characterised.

I keep a small, considered tool surface — Python and PyTorch for daily research, TypeScript and React for web tooling, C and C++ when the budget demands it. I prefer to read fewer papers more carefully than many papers thinly. My research-writing practice leans on FAIR principles and systematic review; the methodology training came through quantitative and qualitative coursework at Dronacharya.

Background#

I am reading B.Tech in Computer Science & Information Technology at the Department of CSIT, Dronacharya Group of Institutions (AKTU Lucknow), 2023–2027. Coursework anchors are machine learning, design and analysis of algorithms, data structures, computer organisation and architecture, and cyber security.

In May 2025 I trained in quantum computing through a one-month programme conducted by CDAC India and IIT Roorkee under MeitY, and followed it with the Qniverse webinar series across the autumn. I ranked first nationally on the CDAC Qniverse examination among more than 3,000 participants (98/100), and I hold the Qniverse Developer certification. In 2024 I was awarded the DronArjun Award for Research, the Odyssey Therapeia Travel Grant for international paper presentation, and admission to the International Association of Engineers (IAENG). My earlier publications cover cloud computing for healthcare interoperability and a 2023 paper on the psychological landscape of AI deployment with collaborators at Dronacharya.

I'm part of Nautilus Lab alongside Rachit Sehgal. The lab works closely with Athena (Odyssey's biology unit) and with Punarbhava (Dr. Roopesh Narayanachary) on the joint medical-device program. Co-authors on recent papers include Roopesh N, Abraham Abel, Neel Raval, Shebin Mohan Kodumthara, Ritu Soryan, and Ishita Soryan.

Conference circuit#

I attend and present at academic conferences regularly — the ICTIS and ICT for Intelligent Systems track in particular, where two of my papers have appeared. In June 2025 I attended the 5th UNESCO– Government of India Stakeholder Consultation on the AI Readiness Assessment Methodology in New Delhi. I have moderated sessions in Pune, Jaipur, Bangkok, London, and New York, and presented in Ahmedabad, Bangkok, Lucknow, and Greater Noida. The full list lives in the CV.

Collaboration#

The fastest channel is the contact form. Email is fine for slower threads. I aim to reply within a week; if I have not, please nudge me.

For research collaboration inquiries, please include in your first message the problem domain, what you have tried, and what you would want from working together. This is more efficient than a back-and- forth and gets to the substantive question faster.

Beyond research#

I read narrative non-fiction (medical history, infrastructure, war reporting). I keep notes on engineering systems that age gracefully — there is a small literature on this and not enough practitioners.