Pranav Garg

AI Researcher

New York, New York, United States16 yrs 11 mos experience

Most Likely To SwitchAI ML Practitioner

Key Highlights

Led GenAI efforts in Amazon Quick Suite.
Published in top-tier ML and PL venues.
Developed innovative algorithms for software verification.

Stackforce AI infers this person is a SaaS expert with a focus on AI-driven software solutions.

Contact

Skills

Core Skills

Ai For CodeAutonomous AgentsApplication SecurityMachine LearningNetwork Security

Other Skills

Memory and PersonalizationSupervised FinetuningLarge Language Models (LLM)AI AgentsNeuro-Symbolic AIGenerative AIPyTorchWord EmbeddingsMXNetAnomaly DetectionTest Case GenerationSoftware FuzzingGNU Compiler Collection (GCC)Computer ScienceAlgorithms

About

I am a Principal Applied Scientist in AWS in the Agentic AI organization. I am currently leading GenAI efforts in Amazon Quick Suite that offers enterprise search, Q&A, data analysis and BI capabilities and recently GA'ed in Oct 2025. Before this, I led efforts in the Q Developer, CodeGuru and AI Security organizations at AWS.My research background spans GenAI, agents, machine learning, formal methods, code analysis, compilers and programming languages. I've published in several top-tier ML and PL venues (NeurIPS, OOPSLA, POPL,ICSE, CAV) and hold multiple patents.Prior to joining Amazon, I finished my PhD from the University of Illinois at Urbana-Champaign (UIUC) on machine learning based automated program verification. In particular, I developed learning algorithms for automatically inferring inductive program invariants that aid software verification.

Experience

16 yrs 11 mos

Total Experience

2 yrs 6 mos

Average Tenure

8 yrs 3 mos

Current Experience

Amazon web services (aws)

3 roles

Principal Applied Scientist

Promoted

Apr 2025 – Present · 1 yr 1 mo

Aug - Till Date
Principal Scientist in Amazon Quick Suite https://aws.amazon.com/quicksuite/
April - July 2025
Science manager in the Code Agents team in Amazon Q Developer, building autonomous agents for code.

AI for codeAutonomous agentsMemory and PersonalizationSupervised Finetuning

Senior Applied Scientist

Promoted

Apr 2020 – Apr 2025 · 5 yrs

Tech lead and science manager in Amazon Q Developer
AI agents for coding tasks including reviewing code, generating documentation and unit tests: https://www.youtube.com/watch?v=se1n10dx1gA&t=493s
Security scanning for bugs: https://aws.amazon.com/blogs/devops/code-security-scanning-with-amazon-q-developer/
Gen AI powered code remediations: https://aws.amazon.com/about-aws/whats-new/2023/11/amazon-inspector-aws-lambda-code-scanning/
detecting hallucinations in AI powered code generations https://arxiv.org/abs/2410.01103 (NeurIps 2025)
Structural code search using NL https://arxiv.org/abs/2507.02107
UTFix: Change Aware Unit Test Repairing using LLM https://dl.acm.org/doi/10.1145/3720419 (OOPSLA 2025)
Scientist in Amazon CodeGuru https://aws.amazon.com/codeguru/ working at the intersection of machine learning and program analysis.
Security code scanning in Amazon Inspector: https://aws.amazon.com/about-aws/whats-new/2023/02/code-scans-lambda-functions-amazon-inspector-preview/
Synthesizing static code analysis rules from code examples: https://www.amazon.science/publications/synthesizing-code-quality-rules-from-examples
https://aws.amazon.com/blogs/devops/resource-leak-detection-in-amazon-codeguru/

Application SecurityLarge Language Models (LLM)AI AgentsNeuro-Symbolic AIAI for code

Applied Scientist

Feb 2018 – Apr 2020 · 2 yrs 2 mos

AISec 2019 paper on a new anomaly detection algorithm we developed for detecting AWS account compromise and customer fraud: https://dl.acm.org/doi/abs/10.1145/3338501.3357368
IP Insights algorithm available in Amazon SageMaker for detecting anomalous usage of IP address: https://aws.amazon.com/blogs/machine-learning/detect-suspicious-ip-addresses-with-the-amazon-sagemaker-ip-insights-algorithm/

PyTorchMachine LearningWord EmbeddingsMXNet

Amazon

Research Scientist

Aug 2015 – Feb 2018 · 2 yrs 6 mos · Bangalore

ML applications to AWS Security including detecting malware from network traffic analysis.

Machine LearningNetwork Security

University of illinois at urbana-champaign

2 roles

Teaching Assistant

Jan 2014 – May 2014 · 4 mos

> One of three TAs of CS428 Software Engineering II (Senior Undergraduate Course).
> Mentored 50+ students towards completion of their Senior Software Projects. The development process was Agile and the projects ranged from mobile apps and web applications to video games.
> Prepared and graded assignments and a final exam. Gave a guest lecture on Software Quality Assurance.

Research Assistant

Aug 2009 – Jul 2015 · 5 yrs 11 mos

Worked on the following broad projects:
● Machine Learning Algorithms for Automated Program Verification
● Design and Analysis of Asynchronous and Distributed Systems
☛ Modeling and Verification of Distributed Key-Value Stores for Eventual Consistency.
☛ Sound and Complete Reduction Scheme for Model-checking Asynchronous and Distributed Systems.
☛ Rebound: A Distributed Scheme for Coordinated Local Checkpointing.
● Automated Verification of Data-Structure Programs

Microsoft research india

Research Intern and Software Consultant

Dec 2012 – May 2013 · 5 mos · Bangalore, India

Data-Race Freedom and Memory Safety in Concurrent Linux Device Drivers.
> Designed and developed new deductive verification technology for concurrent programs.
> Developed static program analysis techniques for proving absence of security bugs, such as buffer overflows and data-race freedom, in concurrent Linux device drivers by automated learning of precise, adequate invariants.

Nec laboratories america

Research Intern

May 2011 – Aug 2011 · 3 mos · Princeton, NJ

Automatic Class Unit Test Generation for C/C++ Programs
> Designed and developed a hybrid unit test generation technique for C/C++ programs that combines random test generation with symbolic execution.
> Showcased a significantly higher code coverage compared to feedback-directed random unit test framework.
> Awarded a US Patent and also published the work in the top-tier software engineering conference ICSE.

Google summer of code

Student Software Developer for GCC

May 2009 – Aug 2009 · 3 mos

Traditional Loop Transformations (with Sebastian Pop, GCC Developer, Free Software Foundation)
> Implemented the loop blocking/strip mining loop transformations in the GCC Compiler.
> Developed code part of GCC version 4.6 and onwards.

Epfl

Research Intern

May 2008 – Jul 2008 · 2 mos · Lausanne, Switzerland

Verified Deadlock-immunity in Large Java Systems
> Verified functional correctness of DImmunix, a deadlock-immunity system software.
> Wrote invariants and method annotations at the code-level in the Jahob proof assistant. The verification conditions were automatically discharged by constraint and logic solvers.
> The system guarantees deadlock immunity in large Java programs with modest runtime/verification overheads.