CV

Software Development Engineer 2

Amazon Lab126

LLM inference on edge hardware and voice assistant features for Amazon smart vehicles.

• Optimized, quantized and deployed LLMs on edge hardware, maintaining 93% accuracy on internal benchmarks. Applied mixed-precision quantization and speculative decoding to accelerate token generation under tight on-device memory and latency budgets. Shipped as part of CES 2026.
• Profiled device-converted LLMs on NPU and CPU targets across multiple automotive chip vendors. Benchmarked latency, memory footprint and accuracy trade-offs to drive model selection decisions for production deployment.
• Created a facial recognition system for in-vehicle profile selection and contributed to in-vehicle LLM trip planning and playlist curation with Claude Sonnet, showcased at CES 2025.
• Developed C++ APIs enabling voice assistant integration with third-party apps for automotive navigation and search, including start and stop navigation, waypoints, and context retrieval.

Software Engineering Intern

VMware Inc.

• Created a C++ data transport plugin for lightweight delta snapshots enabling efficient and scalable VM data protection by backup to cloud providers such as AWS S3. Created design UMLs and test drivers on ESX using Python.

Research Associate

Precog Research Lab, IIIT Delhi

• Developed a full-stack social media analytics tool using Python, JavaScript and MongoDB, containerized with Docker on Linux for 250% faster deployments.
• Built an Ethereum blockchain based decentralized crowdsourcing platform using Solidity and React, with a decentralized reputation scoring system for high quality participation.
• Administered Ubuntu and Linux server machines for lab projects.

Openlab Summer Student (Software Intern)

CERN Openlab

• Developed a consistency verification tool using PL/SQL for metadata databases, reducing redundancies and increasing efficiency for the Compact Muon Solenoid experiment.

Master of Science

Carnegie Mellon University, Information Networking Institute

Information Security

Bachelor of Technology

Maharaja Agrasen Institute of Technology

Computer Science

Running Large Models on Small (Edge) Devices

Poster, Stanford Graph Learning Workshop

Inference optimization for edge model deployment, covering custom KV caching, speculative decoding, mixed-precision quantization, and quantization-aware training, and how they increase accuracy and token generation throughput for edge LLMs.

Speciesism in the Assistant Axis: Probing Compassion Vectors in Post-Trained LLMs

Poster, MLSys Young Professionals Symposium

With Jasmine Brazilek. Extracted assistant, human-compassion and animal-compassion persona vectors from Qwen3-32B served with vLLM on 4xL40S GPUs. Projected activations onto these axes and quantified speciesism between human and animal compassion.

Investigation of Biases in Identity Linkage Datasets

ACM Symposium on Applied Computing (SAC '20)

With Rishabh Kaushal and Ponnurangam Kumaraguru. Machine learning models used to find similarity in large self-collected social media username datasets.

WorkerRep: Immutable Reputation System for Crowdsourcing Platform Based on Blockchain

arXiv

With Gurpriya Kaur Bhatia, Alpana Dubey and Ponnurangam Kumaraguru. A decentralized reputation system for crowdsourcing platforms built on the Ethereum blockchain.

A Survey of Typo-Squatting Attacks in Development Ecosystems

Course project, Network Security, Carnegie Mellon University

A survey of typo-squatting attacks across software development ecosystems. Awarded best paper in the class.

Database Consistency Verification for CMS Metadata

CERN Openlab report, Zenodo

With Valentin Kuznetsov, Katarzyna Maria Dziedziniewicz-Wojcik and Nicolò Magini. A cross-consistency tool providing periodic updates about discrepancies between two CMS metadata databases, DBS and PhEDEx.