CV

Founder and Lead Researcher

Adhara AI Labs

Independent AI research lab

• Founded an independent research lab focused on generative AI, LLMs, compiler optimization, and deep learning systems.
• Led research from concept to production across projects including PyraFuseNet, SLMs as Compiler Experts, RegimeNAS, MorphNAS, and a low-latency jailbreak prevention framework.

Research Assistant (ML Security and LLM Privacy)

Dartmouth College

Advisor: Prof. Shawn Shan

• Co-developed Hierarchical Extraction Search, a query-efficient framework for extracting memorized training data from instruction-tuned LLMs under black-box constraints.
• Demonstrated 10-100x lower extraction cost than brute-force sampling across 12 state-of-the-art LLMs and 100+ attack runs.
• Showed that models appearing robust to high-volume attacks remain vulnerable to low-footprint extraction.

AI Research Intern

Lossfunk

Low-resource language modeling and world models

• Built methods for extremely low-resource Indic language modeling in Tulu, around 0.001 percent of typical training data volume.
• Reduced catastrophic language leakage from 80 percent to 5 percent using hard negative constraints.
• Investigated world-model induction in transformers through modified objectives and architectural biases.

Intent-Based Network Management System (3GPP SON)

Nokia

Industry research internship

• Designed a self-organizing network system that translated high-level operator goals into 3GPP-compliant intents.
• Built an LLM-powered intent classification and orchestration pipeline with conflict detection, duplicate suppression, and intent expiry management.

Contributor

Google Summer of Code 2025, University of California Santa Cruz

Billion-scale ANN benchmarking infrastructure

• Built billion-scale vector embedding benchmarks from open-source codebases for realistic ANN evaluation.
• Developed infrastructure for large-scale evaluation of approximate nearest neighbor algorithms.

Undergraduate Research Intern (Deep Learning)

Indian Institute of Technology, Indore

Advisor: Prof. Nagendra Kumar

• Designed the KASPER framework for PDF malware detection with 99.5 percent accuracy and adversarial robustness.
• Built a malware injection pipeline and explainability analysis using Kolmogorov-Arnold Networks.

Teaching Assistant, Machine Learning

PES University

Teaching and mentoring for undergraduate ML

• Mentored 75+ students across tutorials, labs, and practical assignments.
• Prepared instructional material, delivered tutorial sessions, and conducted weekly hands-on labs.

Subject Matter Expert, Deep Learning

PESU I/O

Deep Learning from Scratch course

• Co-instructed a 30-hour Deep Learning from Scratch course for 40 students.
• Delivered 20 hours of in-person instruction on neural networks, CNNs, RNNs, LSTMs, GANs, and explainable AI.

Head of Technology

Entrepreneurship Club, PES University

Student leadership and technical operations

• Led a 10+ member technical team and managed technology for entrepreneurship events and the club website.

Webmaster

IEEE Robotics and Automation Society, PES University

Student organization web and infra support

• Managed technical infrastructure and web presence for the university robotics organization.

Community Mentor and Technical Speaker

Independent

Mentorship, workshops, and community service

• Mentored 50+ teams in hackathons, research projects, and WiDS Datathon competitions.
• Delivered technical talks on DSPy, LoRA fine-tuning, and deep learning fundamentals.

PhD

Dartmouth College

Computer Science

• Advisor: Prof. Shawn Shan
• Research interests: trustworthy machine learning, AI security and privacy, large language models, memorization and generalization theory, adversarial machine learning, responsible AI systems

BTech

PES University

Computer Science Engineering

• Relevant coursework: Machine Learning, Deep Learning, Generative AI, Linear Algebra and its Applications, Statistics for Data Science, Data Structures and Algorithms, Operating Systems, Computer Networks, Distributed Systems

Resistant to Lawyers, Defeated by Disagreement: Evaluation Blindspots in Legal Language Models

ICML 2026 Workshop on AI for Law

Demonstrated that legal LLMs abandon initially correct answers in 62.1 percent of adversarial conversational interactions.

SLMs as Compiler Experts: Auto-Parallelization for Heterogeneous Systems

NeurIPS 2025 Workshop on Machine Learning for Systems

Achieved a 6.81x average speedup and 43.25x peak speedup over LLVM Polly, TVM, and Triton baselines.

PyraFuseNet: Dual-Path Network for Resource-Constrained Vision

ICIAI 2025

Achieved a 55 percent computational reduction compared with ResNet-18.

GUARDIAN: Multi-Agent Defense System for Large Language Model Security

ICIAI 2025

Proposed a multi-agent architecture with 99.5 percent prompt injection and jailbreak detection accuracy.

Can LLMs Learn Tulu? Teaching Low-Resource Languages Through Hard Constraint Prompting

EACL 2025 LoResLM Workshop

Showed that explicit hard constraints substantially improve low-resource language acquisition in LLMs.

The Prompt Space is Low-Rank: Efficient Memorization Extraction from Instruction-Tuned LLMs

Under review at NeurIPS 2026 Main Conference

Introduces CHASE, a Gaussian-process-based framework that exploits low-rank prompt structure to extract memorized training data from production LLM APIs.

Amazon AI-ML Scholar India

Amazon Web Services

Selected among the top 1,000 AI and ML students in India.

Winner, Cisco ThingQbator Hackathon 6.0

Cisco

National hackathon win among 1,000+ competing teams.

Oxford Machine Learning Summer School

University of Oxford

Accepted to a selective summer program on advanced machine learning.