I'm a Master's student in Computer Science (AI) at Université de Montréal and Mila - Quebec Artificial Intelligence Institute, holding a GPA of 4.3/4.3.
My current work spans two areas. In RL for drug discovery, I am building a discrete masked diffusion model that combines RL-based property optimization and natural language conditioning. In loss of plasticity, I am studying why deep RL agents degrade in learning capacity over time and whether evolutionary-inspired mechanisms can restore it. More broadly, I am drawn to questions about causal structure in RL, particularly whether latent variable models that recover causal environment dynamics could improve agent generalization, and how RL can steer generative models toward open problems in scientific discovery.
Before Mila I spent 4+ years at Deutsche Bank's Chief Innovation Office, working on financial AI for regulatory requirements, LLM agents, fine-tuning LLMs for internal deployment, and safety evaluation of LLM systems. I hold a Bachelor's in Information Technology from NIT Karnataka.
News
Research
Research projects at Mila and selected past work.
-
2026ONGOING
Discrete Diffusion for Molecular Generation with RL and Natural Language Steering
Université de Montréal, April 2026
Learns drug-like molecule representations through discrete masked diffusion over SELFIES, with 100% chemical validity by construction. REINFORCE with per-step gradient accumulation steers generation toward desired properties, improving drug-likeness by 12.8%. Natural language conditioning via SciBERT cross-attention reaches 76% of SOTA at 15% of the parameter count.
-
2025ONGOING
Reinforcement Learning in Non-Stationary Environments: Loss of plasticity and evolutionary remedies
Ongoing research · Mila, 2025–present
Investigating the loss of plasticity phenomenon in deep RL agents and experimenting with evolutionary algorithms to overcome it.
-
2021THESIS
Progressive Conformer for Image Classification
NIT Karnataka, 2021
A hybrid architecture combining Transformers and CNNs at the attention level for progressive feature learning in image classification.
Projects
Selected engineering and applied ML projects.
Expected Goal Prediction · NHL
Goal probability prediction for NHL gameplay. Peak AUC 0.88. Engineered 18+ spatial/temporal features; Bayesian and Grid Search hyperparameter tuning.
AI Agents for Computer Use
LLM agent that browses the web via natural language instructions, using DSPy for few-shot prompting over cleaned HTML context.
Deep UNet · Melanoma Segmentation
Image segmentation model for dermoscopy images used in downstream disease classification tasks.
Light-to-Camera Indoor Positioning
Novel indoor positioning system for mobile devices using light-based communication.
AI for Dementia
iOS app helping dementia patients manage routines via voice-to-text and text-to-voice models.
More on GitHub.
Skills
Education
M.Sc. Computer Science, Artificial Intelligence
Université de Montréal · Mila, Quebec AI Institute
GPA 4.3 / 4.3 · Fundamentals of ML, Data Science, Representation Learning
B.Tech. Information Technology
National Institute of Technology Karnataka (NITK), India
CV, Neural Networks, Deep Learning, Algorithms, Distributed Systems, HPC, Graph Theory
Experience
Research Intern, LLM Agents and Software Co-pilot
CRIM (Centre de Recherche Informatique de Montréal) · Montréal, QC
Research on enabling LLMs to navigate and operate complex software as genuine co-pilots for end users, applying RL, supervised fine-tuning, and RLVR to train agents for real-world application use.
Associate / Senior Analyst / Analyst, Chief Innovation Office
Deutsche Bank Group · Pune, India
Financial AI for regulatory requirements, LLM agents for internal use cases, fine-tuning LLMs for internal deployment, and safety evaluation of LLM systems.
Software Engineer Intern
Pulse Secure LLC · Bengaluru, India
Deep LSTM for insider-threat detection; Node.js log visualisation platform.