I recently graduated from M.S. Ramaiah Institute of Technology, Bangalore, with a degree in Electronics and Communication Engineering and a minor in Computer Science.
For 18 months I was a Research Intern at the SCALE Lab, Department of Design and Manufacturing, Indian Institute of Science (IISc), Bangalore, under Sankar Balasubramanian.
My research centred on making AI useful inside the designer's actual workflow — generating visual inspiration on demand, tracking design evolution through version control, and classifying designer intent in real time. The broader question driving the work: how do you reduce the cognitive overhead of creative decisions without replacing the designer's judgment?
My broader interests are in deep learning, computer vision, and the mathematics underlying both. I am currently applying to MS programs in AI/ML for Fall 2027.
Education
M.S. Ramaiah Institute of Technology, Bangalore
B.E. Electronics and Communication Engineering
Minor in Computer Science
Vidya Mandir Independent PU College, Bangalore
Pre-University (Science)
Research
D2 Design: Dual-conditioned Diffusion AI for Accelerating Product Form Generation
Computer-Aided Design, 2026 · FormForge Tex · FormForge Img
Traditional product form development is slow and expertise-dependent, progressing through mood boards and sketches before a single rendering exists. We present FormForge, an AI platform built on fine-tuned Stable Diffusion models conditioned on text descriptors and hand-drawn doodles. In a formal study, FormForge reduced time-to-concept by 85% (text) and 62% (sketch), lowered NASA-TLX cognitive load scores by up to 33 points, and produced the preferred design in all three product categories as judged by 25 design experts.
Git for Sketches: An Intelligent Tracking System for Capturing Design Evolution
Conventional sketching tools discard the non-linear history and cognitive intent behind design decisions. DIMES introduces sGIT, a visual version control system that maps Git primitives to design actions and captures stroke data alongside voice intent. In a comparative study, DIMES users showed a 160% increase in concept exploration breadth, and AI-generated narrative summaries enabled novices to replicate designs with significantly higher fidelity (cosine similarity 0.97 vs. 0.73).
Robust Multimodal DeepFake Detection Using Transfer Learning and Feature Engineering
Submitted and accepted to the 16th Intl. Conference on Computing, Communication and Networking Technologies (ICCCNT)
A dual-approach DeepFake detection framework combining ResNet-50 transfer learning with handcrafted numerical and wavelet-based features for ML and DL classifiers. On the DF-WILD CUP dataset, the hybrid approach achieves up to 93.5% accuracy.
Projects
Tvacha — Skin Health App
Deep LearningCollaborated with a dermatologist to design a custom CNN for multi-class skin cancer detection. Incorporated domain-specific medical markers into the model pipeline, achieving 94% classification accuracy on curated dermatological datasets.
Dark Patterns Detector
LLM · Chrome ExtensionA Chrome extension that identifies deceptive UI patterns — false urgency, subscription traps, basket sneaking — on e-commerce sites. Encodes page screenshots in Base64 and runs them through a fine-tuned IDEFICS-9B vision-language model, producing annotated overlays directly in the browser.
Working at DNV
Aug – Oct 2024Monitored and analysed data from renewable energy assets to optimise performance. Built an interactive GUI in Tkinter and PyQt5 for data visualisation, and processed large CSV datasets with Pandas for real-time monitoring of energy resources.
Working at Li2 Technologies
Sep – Oct 2023Built an interactive learning platform for personalised education in underprivileged classrooms, using Python, LangChain, and NeMo Guardrails to ensure safe, contextually relevant AI responses. Delivered as part of a multi-member team.