Conducted security research on agentic AI systems, synthesizing 10+ foundational papers into a unified threat model covering prompt injection, memory poisoning, credential leakage, and unauthorized autonomous actions. Performed red-team analysis of OpenClaw agents, reproducing real-world incidents such as the Shellraiser token launch and malicious agent tooling, and translating observed failures into concrete security tests and mitigations.

Conducted quantitative startup analysis for a student-run venture capital fund, sourcing and evaluating 40+ early-stage fintech, AI, and SaaS companies using data-driven market research and competitive analysis. Built 3-statement financial models, unit economics, and DCF valuations for 10+ startups, applying KPI benchmarking, growth decomposition, and sensitivity analysis to support high-conviction investment decisions.

Conducted research on data-driven biometric cryptography solutions, co-developing Face Recognition Privacy models with 92% accuracy using ResNet, DenseNet, and SVMs. Engineered CUDA-accelerated feature extraction algorithms reducing runtime by 40% while processing 400K+ structured and unstructured samples. Documented ML architectures achieving 90–94% accuracy, enhancing data communication and automation.

Automated Python-based data collection and visualization workflows for high-precision laser calibration experiments. Applied statistical regression models to improve measurement accuracy and instrument control. Supported demonstrations and reports for 50+ researchers, improving productivity and data organization within the research team.

Engineered a modular robotics software library with intuitive user interfaces and scalable architecture. Collaborated with the robotics team to translate functional requirements into efficient control algorithms. Authored documentation ensuring maintainability and extensibility for future teams.