An AI agent that traces a single entity's evidence across graph, relational, and text stores into a cited analytic note. Built on the Claude Agent SDK.
Projects
Designed and delivered.
Interactive multi-agent development system: ten specialized agents stream their reasoning, ask for guidance, and iterate live across terminal, pygame GUI with TTS, and a Ren'Py visual novel.
Rust + Python federated learning orchestrator with paper-cited aggregation strategies and round-level / data-pipeline attack simulations.
Windows security assessment tool that scores system metrics (patches, ports, firewall, antivirus) via Python rules, a CLIPS expert system, ML neural networks, or federated learning.
Federated LoRA fine-tuning testbed on the current Flower app model, with OpenTelemetry traces and metrics for every round and every client.
Federated personalization testbed whose synthetic reports carry real classification levels and compartments, so two holders can be incomparable rather than ranked. It measures what a personal/shared boundary costs and how much privacy leaks past it under secure aggregation and audit policy.
Claude Code plugin marketplace of agent workflows for repository maintenance, adversarial code review, LaTeX and slide builds, and bibliography verification against arXiv and Crossref.
Webots robot that learns to reach goals and avoid obstacles, with distance-adaptive motor control. A tabular Q-learning agent and a PyTorch Double DQN (target network, replay buffer, early stopping) learn from the same state encoding and reward, which includes Ivan Rojas's potential-based reward shaping.
Federated learning framework for configuring, benchmarking, and comparing FL strategies under static and dynamic adversarial attacks, with pluggable robust aggregation and an experiment-planning AI agent.
Public website for the Laboratory of Data Quality and Intelligent Security at RIT, live at dataqualitylabs.com. An Astro and Tailwind build with a git-backed CMS replaces the lab's legacy Flask + Bootstrap site.
Mobile robot control systems using Python and Webots covering localization, navigation, and sensor integration.
Multi-domain support-triage AI agent on Claude Opus 4.7: prompt caching, structured tool-call output, prompt-injection defense, and Anthropic Message Batches API. Built solo for HackerRank Orchestrate (May 2026).
Real-time biosignal sonification: EMG/EEG from the BioRadio streams over Lab Streaming Layer and maps to MIDI notes for live performance. AWARE-AI Spring Hackathon team project with Victor Lockwood, Parth Kapur, Grant Bosworth, and Sophia Caruana (Feb 2026).
Private blockchain system and admin dashboard for DocBridge.io document certification: Proof-of-Work mining with SHA-256, transaction explorer, user management, Google OAuth 2.0 + JWT. Senior capstone (USF, May 2022).