Portfolio
A collection of research tools, engineering platforms, and personal explorations — each built to solve a real, well-defined problem.
Research & Personal Projects
A GPU-accelerated protein stability prediction engine implementing a custom molecular mechanics force field. Predicts the energetic impact of amino acid mutations using AMBER ff14SB parameters, LCPO solvation, and Numba/CUDA acceleration. Validated against FoldX and experimental benchmarks on p53 (1TUP). Includes a FastAPI REST microservice and Kubernetes deployment manifests.
A full-stack Instagram Data-as-a-Service (DaaS) intelligence engine built from scratch. Combines a stealth Playwright-based scraper, SQLite database, multi-page Streamlit analytics dashboard, and a custom Chrome Extension (MV3) to extract and monetise Instagram data — solving 20 major engineering blockers across Meta's security architecture, React 18 internals, and Chromium browser limits.
A high-performance bridge between the Geant4 high-energy physics simulation toolkit and Blender's 3D rendering engine, communicating via gRPC and Protocol Buffers. Enables real-time, interactive particle physics visualisation with <100ms latency, a single-mesh BMesh architecture for 10-50x viewport performance, and an interactive run control state machine.
A research-grade, zero-false-positive static analysis engine for Python. Unlike standard linters, OptiScan targets "silent performance killers" — code patterns that are syntactically valid but computationally catastrophic. Built on LibCST with AI-native JSON reporting designed for direct consumption by autonomous LLM-based refactoring agents.
An AI-ready code intelligence and visualisation platform that transforms complex Python repositories into interactive, human-readable knowledge graphs. Uses a custom AST processing engine for semantic decomposition and a physics-based force-directed graph frontend (vis.js) to reveal hidden dependencies, cyclomatic complexity hotspots, and architectural bottlenecks.
A professional-grade computational suite for high-precision astronomical mechanics. Implements 50+ specialized Python modules covering recursive multi-level timing algorithms, harmonic divisional systems, and interactive 3D celestial modeling. A demonstration of quantifying complex rule-based knowledge systems through rigorous software architecture.