Education
University of California, Davis
Bachelor of Science
Computer Science and Engineering
2019 - 2023
Work Experience
Amazon
Software Development Engineer 2 (SDE 2)
July 2023 - Ongoing
Sunnyvale, CA
- Spearheaded infrastructure for World Catalog, Amazon's pioneering agent-first global catalog, harmonizing non-Amazon product data for next-gen AI search. Architected the Progressive Listing Experience (PLE) using LLMs to autonomously map seller URLs to Amazon schema. Led a ground-up rewrite of the ingestion pipeline for async LLM workflows, enhanced system traceability, and optimized CI/CD pipelines to reduce build times by over 80%. Launched 3 microservices with a plugin architecture and 200+ custom rules for data validation and image enrichment.
Amazon
SDE Intern, Item Pipeline
June 2022 - Sep 2022
Sunnyvale, CA
- Engineered a custom Code Analyzer to automate bug detection and enforce coding standards, adopted by 100+ engineers across 8 teams. Architected and deployed scalable serverless microservices for the Item Pipeline utilizing Java, TypeScript, AWS Lambda, Amazon SNS, and VPCs.
Opal, LLC.
Junior Developer
Sep 2021 - June 2022
Remote
- Engineered a scalable online NFT marketplace utilizing Flask, Django, and React.js. Spearheaded the Social Media team, leading the end-to-end development of an interactive platform for user engagement and digital asset sharing.
Google Developer Student Club, UC Davis
Web Developer, Tech Lead
February 2020 - July 2022
Davis, CA
- Architected the official GDSC UC Davis website using dynamic JavaScript, HTML, and CSS. Collaborated on the development of a student-led Kotlin Android development curriculum.
Bit Project
Workshop Lead
Oct 2019 - Feb 2020
Davis, CA
- Partnered with engineering teams at MongoDB, Pusher, and Twitter to develop and deliver technical workshops to 50+ attendees. Authored comprehensive technical tutorials focusing on the MERN stack and complex API integrations.
Niyantras
Intern
May 2019 - Aug 2020
Vadodara, IN
- Engineered a cost-effective, BLE-based IoT solution by integrating Raspberry Pi modules with a custom mobile application for real-time hardware interaction.
Projects
- Engineered the scalable core infrastructure for an evolutionary AI red-teaming framework and fuzzer, optimizing high-throughput execution pipelines for adversarial prompt injection. Architected data systems to support real-time attack mutation using genetic algorithms. Won 2nd place out of 1,500+ submissions at the Berkeley RDI AgentX competition; framework later forked by Mozilla's 0din team.
HASHIRU
Jan 2026 - March 2026
PythonLangGraphAgentic AIGenAIGradioHugging Face
- Architected a budget-aware multi-agent orchestration framework (Hierarchical Agent System for Hybrid Intelligent Resource Utilization) enabling highly distributed workload management. Built robust cost-optimization controls to strictly manage resource compute constraints, inference expenses, and token budgeting across parallel LLM agent execution streams.
- Engineered advanced search functionality and a comprehensive categorization system for the gem5 Resource Infrastructure, enhancing resource discovery for computer architecture researchers. Architected extensible database integrations to support diverse storage and retrieval options, providing developers with robust tools for managing simulator resources.
- An open-source, automated portfolio builder that allows users to deploy professional portfolios to GitHub Pages in minutes. Developed a template-driven engine that achieved over 1,000 page views in its first week of alpha launch.
- Reverse-engineered and bypassed Google Stadia Controller hardware constraints, implementing a websocket-based architecture to enable universal wireless connectivity. Accumulated 450+ stars on GitHub and featured in major publications including The Verge and Ars Technica.
- A diet-restriction assistant app that identifies safe-to-eat products by scanning ingredient images. Awarded 'Most Social Good Hack' at HackDavis and selected as a Top 20 global solution in the Google Solution Challenge.
Publications
NAAMSE: Evolutionary Security Evaluation Framework for Agents
paper
Parth Shah, Harshil Patel, Kunal Pai
arXiv / Berkeley RDI AgentX
Introduces NAAMSE, an evolutionary AI red-teaming framework and fuzzer designed to test LLM agents for security and utility. The framework utilizes genetic algorithms for real-time attack mutation, balancing Attack Success Rate (ASR) and False Positive Rate (FPR) to mitigate the 'over-blocking' problem in agentic systems. Recognized with 2nd place at the Berkeley RDI AgentX competition.
AI SecurityRed-TeamingAgentic AIGenAI
HASHIRU: Hierarchical Agent System for Hybrid Intelligent Resource Utilization
paper
Kunal Pai, Parth Shah, Harshil Patel
arXiv
Rapid Large Language Model (LLM) advancements are fueling autonomous Multi-Agent System (MAS) development. However, current frameworks often lack flexibility, resource awareness, model diversity, and autonomous tool creation. This paper introduces HASHIRU (Hierarchical Agent System for Hybrid Intelligent Resource Utilization), a novel MAS framework featuring a 'CEO' agent dynamically managing specialized 'employee' agents. Evaluations demonstrate HASHIRU's capabilities in complex reasoning, outperforming Gemini 2.0 Flash on benchmarks like GSM8K (96% vs. 61%) and JEEBench (80% vs. 68.3%).
Agentic AIGenAIMultiAgent Framework
gem5 Vision
poster
Parth Shah, Kunal Pai, Harshil Patel, Arslan Ali
ISCA 2023: gem5 Workshop
The gem5 Vision Project seeks to improve user-friendliness and accessibility by introducing advanced search functionality, comprehensive resource categorization, and expanded database support within the gem5 ecosystem for researchers and developers.
Computer Architecture
Awards
Amazon
Bias for Action Award
2024HackDavis
Best Use of GitHub
2023HackDavis
1st Place
2022SacHacks
Best Local Hack
2022HackDavis
2nd Place
2021PennApps XXI
Top 30
2020Google Solution Challenge
Top 20
2020HackDavis
Best Social Good Hack, Best Use of Google Cloud
2020