PRATHAM PATEL - TECHNICAL BLOG
About Pratham Patel
AI/ML Engineer & Full-Stack Developer
[ SCROLL_TO_EXPLORE ]
My Journey in Tech
As a Computer Science student at Gannon University, I've dedicated myself to exploring the intersection of artificial intelligence and software development. My journey began with a fascination for how machines can learn and adapt, leading me to specialize in reinforcement learning and natural language processing.
From developing adversarial robustness models for Android malware detection achieving 97% accuracy, to architecting novel LLM reasoning frameworks with 60% performance improvements, I'm driven by the challenge of pushing technological boundaries while creating practical solutions.
As the Founder and President of the Gannon Codex Programming Club, I've grown our community to 50+ members, organizing AI/ML workshops and hackathons. My role as a CIS Lab Technician has taught me the importance of maintaining robust systems, achieving 95% lab uptime and reducing ticket resolution time by 30%.
"Technology is a canvas. I use it to explore ideas, solve problems, and build experiences. But beyond the code, there's a human story."
[ TOOLKIT ]
[ LANGUAGE_PROFICIENCY ]
[ 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.
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.
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.