Loading models, metrics, and questionable life choices...
I’m an AI/ML researcher and product builder turning language, vision, and time-series models into tools people can use.
Currently reseraching on how Prompt Injections can affect persistant memories in LLMs at FLAME University.
Research depth and working systems, with outcomes drawn from the latest résumé.
A production-facing automated essay-scoring workflow combining regression models, text embeddings, grammar checks, sentiment analysis, AI detection, and a summary dashboard.
A Per-Dimension Top-K pooling strategy for BERT that improves resistance to adversarial token injection without changing the underlying model architecture.
An OCR-assisted text-image classification system that labelled promotional, neutral, and advisory content at research scale.
A VAE and CNN pipeline designed to improve cancer detection on imbalanced histopathology datasets.
Explore the technical, research, and delivery skills behind the work.
Applied research grounded in measurable operational outcomes.
Not All Tokens Are Equal: Per-Dimension Top-K Pooling for Adversarially Robust BERT Classification
ACL GEM 2026Unveiling Smoking Behavior Dynamics Through Graph Neural Network Analysis
Springer LNNS
I earned a B.Tech in Electronics and Communication from VIT Vellore. My work sits between data science and product: building AI systems that make complex workflows faster, clearer, and more useful.
For research, product, and applied machine-learning opportunities.