Praveenkumar Kanithi

Clinical AI Researcher, University of Oxford

About

Hi, I’m Praveen, a Senior Researcher at the University of Oxford, working in healthcare AI. I work on machine learning for healthcare, across areas like clinical language models, genomics, electronic health records, and medical imaging. My day-to-day mostly involves figuring out how to make models a bit more useful, reliable, and helpful in real-world clinical settings.

I did my PhD at the HITLab in New Zealand, and masters from IIT Kharagpur. Before moving to the UK, I spent close to 6 years at M42 in Abu Dhabi as a Principal Scientist. Outside of work, I enjoy hiking, traveling, and training Brazilian Jiu-Jitsu.

Praveenkumar Kanithi

Projects

Background

Work Experience

Senior Researcher – Healthcare AI

University of Oxford, UK

03/2026 – Present

Principal Scientist

M42 (A G42 company), Abu Dhabi, UAE

09/2023 – 03/2026

Senior Applied Scientist

M42 (A G42 company), Abu Dhabi, UAE

06/2020 – 09/2023

Education

Ph.D.

HITLab, University of Canterbury

03/2017 – 05/2020

Thesis: Interactive image segmentation of MARS spectral CT datasets

This thesis addressed the challenge of segmenting spectral CT images for pre-clinical studies, where limited datasets and variable imaging conditions make automatic methods unreliable. An interactive segmentation framework was developed, beginning with a bag-of-features approach for 2D images and later extended to incorporate spatial user cues and 2-stage conditional random fields to improve segmentation accuracy at boundaries. The work was further adapted to volumes through a slice-wise strategy, enabling user interactions to propagate consistently across slices. These methods were implemented as software tools to support researchers and clinicians.

M.Tech. in Medical imaging and informatics

IIT Kharagpur

07/2014 - 05/2016

Thesis: Immersive augmented reality system for assiting needle positioning during ultrasound guided intervention

This thesis developed an augmented reality system to support ultrasound-guided needle interventions, where accurate in-plane alignment of the needle with the ultrasound probe is often difficult. The system tracked both the probe and needle in a 3D world coordinate system using fiducial markers, enabling real-time visualization of the projected needle trajectory on the ultrasound image before insertion. The approach aimed to reduce the need for multiple punctures during free-hand procedures and to improve clinician navigation. The visualization was further integrated into a head-mounted display, providing an immersive AR interface that overlaid navigation assistance directly on the live ultrasound feed.

B.Tech. in Electronics and communication engineering

RGUKT Basar

07/2010 - 05/2014

Selected Publications

A full list of publications can be found on Google Scholar. * indicates equal contribution.

Medvedev, A.*, Viswanathan, K.*, Kanithi, P.*, Vishniakov, K., Munjal, P., Christophe, C., ... & Khan, S. (2025). BioToken and BioFM-Biologically-Informed Tokenization Enables Accurate and Efficient Genomic Foundation Models. Accepted at ICML 2026. PDF | Code | Model

Raha, T., Christophe, C., Saadi, N., Javed, H. A., Pimentel, M. A., Rajan, R., & Kanithi, P. (2026). Cross-Examination Framework: A Task-Agnostic Diagnostic for Information Fidelity in Text-to-Text Generation. Accepted at ACL 2026. PDF | Code

Vishniakov, K., Viswanathan, K., Medvedev, A., Kanithi, P., Pimentel, M. A., Rajan, R., & Khan, S. Tokenization to Transfer: Do Genomic Foundation Models Learn Good Representations?. The Fourteenth International Conference on Learning Representations (ICLR 2026). PDF | Code

Maslenkova, S., Christophe, C., Pimentel, M. A., ... Kanithi, P. Building Trust in Clinical Analysis: Bias Analysis and Dataset Transparency. EMNLP 2025. PDF | Dataset

Al-Mahrooqi, A., Munjal, P., Rajan, R., Pimentel, M. A., Kanithi, P. (2025). Empirical Analysis of Scaling Vision Foundation Models for Chest X-rays. In Medical Imaging with Deep Learning. PDF | Model

Kanithi, P. K., Christophe, C., Pimentel, M. A., Raha, T., Saadi, N., Javed, H., ... & Khan, S. (2024). MEDIC: Towards a Comprehensive Framework for Evaluating LLMs in Clinical Applications. arXiv preprint arXiv:2409.07314. (Presented at NVIDIA GTC 2025) PDF | Demo

Christophe, C., Raha, T., Maslenkova, S., Salman, M. U., Kanithi, P., Pimentel, M., & Khan, S. (2024, November). Beyond Fine-tuning: Unleashing the Potential of Continuous Pretraining for Clinical LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2024 (pp. 10549-10561). PDF

Christophe, C., Kanithi, P. K., Munjal, P., Raha, T., Hayat, N., Rajan, R., ... & Khan, S. (2024). Med42--Evaluating Fine-Tuning Strategies for Medical LLMs: Full-Parameter vs. Parameter-Efficient Approaches. AAAI Clinical LLM symposium. PDF | Model