Hello, I'm

Ben

Graduate Researcher & Software Engineer specializing in AI, full-stack development, and systems programming.

Ben

About

I'm a Computer Science graduate student at Stony Brook University, researching Small Language Models and context-efficient reasoning strategies. My interests lie at the intersection of machine learning and full-stack engineering—I enjoy building AI-powered applications that solve real-world problems, from RAG systems to intelligent financial tools.

On the ML side, I work with language models, fine-tuning transformers, and designing efficient reasoning frameworks. In full-stack development, I build scalable web applications using modern frameworks like Next.js, TypeScript, and cloud infrastructure.

  • Python
  • TypeScript
  • Next.js
  • React
  • Node.js
  • FastAPI
  • AWS
  • Docker
  • PyTorch
  • C++

Projects

TrackMyMoney

AI-powered budget tracker with PDF/CSV parsing and automatic categorization. Reduced dashboard loading time by 40% through independent widget loading and async processing with Celery.

DMRAG

Retrieval-Augmented Generation system that automates the Dungeon Master role in D&D. Features hybrid retrieval (BM25 + semantic similarity) and fine-tuned transformers.

99 - Multiplayer Card Game

Real-time multiplayer game with turn-based state management. Features CI/CD pipeline with GitHub Actions, multi-container deployment on AWS EC2, and Redis for low-latency synchronization.

xv6 OS Extensions

Extended xv6 OS with custom system calls for process tracing and dynamic priority management. Implemented FCFS and Priority-Based Scheduling algorithms.

Adaptive Difficulty-Driven Router

Designed an adaptive routing system using Random Forest stacked with XGBoost that selects between Phi2 (lightweight, 41% accuracy) and WizardMath (powerful, 88% accuracy) for GSM8K math problems. Optimized efficiency by routing only 63% of problems to the expensive model while achieving 74.2% overall accuracy.

RL Snake

Implemented a Deep Q Network (DQN) reinforcement learning agent for Snake game that achieved scores of 100+, 400% higher than common implementations. Developed multiple exploration strategies with variable reward systems and collision detection mechanisms, experimenting with state representations and reward engineering techniques.

Unsupervised Domain Adaptation

Engineered a Domain-Adversarial Neural Network (DANN) to mitigate domain shift, utilizing a Gradient Reversal Layer (GRL) to force feature alignment between labeled source and unlabeled target datasets.

Bitcoin Simulator

Designed a Bitcoin simulator in C, implementing a blockchain from scratch using linked lists to chain transaction blocks. Integrated hash tables and dynamic arrays for efficient user data management, balance tracking, and transaction histories. Developed attack simulation and validation mechanisms to ensure chain integrity.

Experience

  • Aug 2025 — Present

    Graduate Researcher

    LUNR Lab @ Stony Brook University

    Investigating context-efficient reasoning strategies for Small Language Models (7-14B) on the AppWorld benchmark. Implemented the TAOS framework to replace full-history retention with iterative state summarization.

  • Aug 2023 — Dec 2023

    Undergraduate Researcher

    SERC Lab @ IIIT-H, Hyderabad, India

    Worked with Prof. Karthik Vaidhyanathan to build a multi-objective reinforcement learning framework to autonomously manage web server scalability. The system dynamically adjusts server resources in real-time to minimize response time and operational costs while maximizing user experience. Implemented using Q Learning and Proximal Policy Optimization (PPO).

  • May 2023 — Aug 2023

    Software Engineering Intern

    NeoBase, Dubai, UAE

    Translated Figma prototypes into responsive Next.js and Tailwind CSS interfaces for Web3 client projects. Maintained production-grade code quality with modular, type-safe TypeScript components.

  • Nov 2022 — May 2023

    Software Engineering Intern

    Inclusive Growth Chain, Hyderabad, India

    Architected the company's official landing page and revamped legacy React.js codebase for a comprehensive portal serving farmers in rural India. Integrated 25+ APIs for loan processing and crop advisories.

Education

  • Aug 2024 — May 2026

    Master of Science in Computer Science

    Stony Brook University, New York

  • Sep 2020 — May 2024

    Bachelor of Technology in Computer Science

    IIIT-H, Hyderabad, India

Contact

I’m open to collaborations and new opportunities.