PRATHAM

PRATHAM

background wave

YOU WANT TO KNOW MORE ABOUT ME?

[ SCROLL_TO_EXPLORE ]

"Technology is a canvas. I use it to explore ideas, solve problems, and build experiences. But beyond the code, there's a human story."

[ PEEK_INTO_MY_LIFE ]

[ RESEARCH_LOGS ]

Synergistic Self-Correction for LLM Reasoning

Architected a novel reasoning framework augmenting LLMs with Proximal Policy Optimization (PPO) and RAG-based grounding to ensure factual consistency. Demonstrated a 60% relative improvement on the GSM8K benchmark.

NLP + RL
PPO

Adversarial Robustness in Android Malware Detection

Constructed a hybrid malware detection model combining static opcode analysis and dynamic runtime behaviors, achieving 97% accuracy on a dataset of 100,000+ APKs. Accepted for presentation at the Microsoft Future Tech Conference.

Cybersecurity
Adversarial AI

Reproducible RL Research Pipeline

As an AI Research Intern at DA-IICT, I engineered a complete RL research pipeline using Docker, reducing model evaluation time by 40% and boosting accuracy by 20% through robust experiment harnesses.

Docker
W&B