projects
hackathons and lab builds, one at a time.
Autonomous Red Teaming Agent
An agent that pentests other agents — hunting prompt injection, jailbreaks, and guardrail bypasses on a loop, then filing structured findings.
Built for the KnowBe4 AI Hackathon as part of the winning team. The system acts as an autonomous red team: it plans attack campaigns, probes target agents with adaptive jailbreak strategies, scores guardrail failures, and reports reproducible traces. The same ideas later informed production agent-security work.
Matrix
Data enhancement for training recommendations and phishing simulations, built for new users with no previously known profile or behavioral history.
Matrix solves the cold-start problem in security awareness. It enriches sparse user data so brand-new people still get personalized training and realistic phishing simulations.
AURATER — Trash-Eliminating Rover
An autonomous unmanned rover that detects and collects trash. This is the project that started the applied-AI streak, published as primary author.
Primary author of “AURATER: Autonomous Unmanned Rover with Advanced Trash Eliminating Robot,” presented at the International Conference on Recent Advances in Engineering and Technology. Computer vision plus mechatronics: detect debris, plan a path, pick it up.
IllustratED
History, science, and math for kids, explained with AI-generated illustrations so the idea lands as a picture, not a paragraph.
Lessons are paired with on-the-fly illustrations so abstract ideas become visual stories. Built to make learning feel like opening a picture book that writes itself.
WhiZED · Socratic Tutor
A Socratic tutor for school students that asks before it answers, built around Gemma 3n so the dialogue can live closer to the device.
Instead of dumping answers, it walks the student through questions, hints, and worked examples. A companion experiment, WhiZardLM, is a workspace for school notes. The tutor itself is WhiZED.
Juno & Asimov
Voice-first English practice for kids, including under-privileged learners, with conversation instead of flashcards.
Juno is an AI spoken-English app. Asimov is a GPT-powered English learning companion. Both explore patient, voice-first practice loops rather than multiple-choice drills.