Human
University of Galway
Jan 2026 – Present · Ireland
Worked on integration of neural networks and computational psychiatry; paper currently under review. Currently researching supply chain attack prevention in software development.
Aug 2025 – Dec 2025
Built FasTest, an AI-powered API testing platform. Engineered multi-shot LLM prompt chains with structured output constraints to auto-generate boundary-aware, adversarial edge cases with minimal human input. Developed a FastAPI async backend and Next.js frontend with real-time test visualization.
2024 - 2028
B.Tech, Information Technology
2024 – Present
CGPA: 9.5/10.0
Built a 107M parameter small language model using a linear attention architecture combining FAVOR+ (Performer) with asymmetric KV projections for sub-quadratic attention. Achieved 23% faster inference than GPT-2 at comparable parameter count; quantized model ran 84% faster. Tech stack: Python, PyTorch.
Built a neural network framework from scratch with a custom Tensor class, reverse-mode autograd via topological sort, and a dynamic computation graph. Implemented SGD with Momentum and Adam optimizers from scratch with full forward/backward propagation. Documented with a live API reference deployed on GitHub Pages. Tech stack: Python.
Merged contributions to pytorch/pytorch, scipy/scipy, and alexandria3k, including scalar NaN propagation in tensor operations, worked examples for the special.roots jacobi Gauss-Jacobi quadrature API, and lookup complexity improvements. Open pull requests include ISSN value space and time complexity improvements and an author blocking feature for a disambiguation pipeline.