Projects
Industry Projects & Launches — Amazon (2024–present)
Focus Areas
- Large-scale synthetic data generation for LLM training and evaluation (e.g., reasoning traces, agentic tool-use benchmarks)
- LLM post-training: SFT and RL (e.g., PPO, GRPO, DAPO, GSPO, CISPO and their variants), including agentic long-horizon RL, for coding and ads agents
- Self-evolving agents for agent harness evolution and LLM post-training as an automated AI researcher
- Coding agents for large-scale code decomposition, migration, documentation, and evaluation
Launched Products I Contributed To
AWS CodeLLM Team
- Amazon Q Developer transformation for .NET, mainframe, and VMware workloads — generative AI agents that streamline large-scale code transformation, supervised by teams through a unified web experience. [AWS News Blog]
- AWS Transform (generally available) — agentic AI that accelerates enterprise migration and modernization. [AWS Blog]
Amazon Ads Foundation Model Team
- Ads Agent (introduced at unBoxed 2025) — agentic AI for simpler, higher-performing advertising. [Amazon Ads]
- Amazon Ads Agent — one AI-powered ad platform bringing together Full-Funnel Campaigns, DVA+, and an enhanced conversational experience. [Amazon Ads]
- Conversational experience in Amazon Ads Agent — natural conversation to get answers, make decisions, and take action across advertising tasks and workflows. [Amazon Ads]
PhD Research Projects — Georgia Tech (2019–2024)
- Large Language Models for Controllable Text Generation via Reinforcement Learning
- Advisors: Prof. Mustaque Ahamad and Prof. Srijan Kumar
- Research output: ACM WWW’23 paper, code and data
- Designed novel BERT-based reward models in an offline RL framework to capture desirable textual properties in text generation
- Implemented a transformer decoder-based GPT-style language model for counter-misinformation generation
- Curated high-quality misinformation-related conversational datasets and fine-tuned LLaMA2 using LoRA
- Multi-task Learning-based Personalized Text Generation for Attack/Defense
- Advisors: Prof. Mustaque Ahamad and Prof. Srijan Kumar
- Research output: ACM KDD’21 paper, code and data
- Proposed a Generative Adversarial Network (GAN)-based text generation framework for personalization and adversarial attack
- Designed a multi-stage multi-task optimization strategy to improve the generation performance
- Leveraged adversarial machine learning for fine-tuning and enhancing the model robustness
- BERT-based Fine-grained Text and Textual Sequence Classification
- Advisors: Prof. Mustaque Ahamad and Prof. Srijan Kumar
- Research output: IEEE BigData’20 paper, code and data and ACM ASONAM’21 paper, code and data
- Implemented masked language model-enhanced fine-tuning- and feature-based BERT for misinformation and hate speech classification
- Proposed a transformer-based sequence-aware deep sequence classifier powered by data augmentation from OpenAI ChatGPT and GPT-4
- Crawled large-scale tweets and user networks, and analyzed the spread of misinformation and hate speech on Twitter