Mikiyas Zenebe

Problem: Legacy surveillance systems show video, but not operational insight.

Solution: Built a real-time face recognition and alerting system for modern campus monitoring.

Impact: Turns CCTV streams into actionable identity, access, and safety signals.

Problem: AI agents need a safety and policy layer before they can operate in real systems.

Solution: Shaped an SDK concept for securing agent behavior, decisions, and execution boundaries.

Impact: Positions Megent as infrastructure for trustworthy agent workflows.

Problem: Classroom coordination still depends on manual updates and scattered communication.

Solution: Built repIt as an AI classroom representative agent to organize messages, reminders, and coordination.

Impact: Gives classroom communication a reliable agent layer.

Problem: Berta-English translation has limited high-quality machine learning support.

Solution: Trained a low-resource translation pipeline for Berta language preservation and access.

Impact: Supports indigenous-language digitization and future NLP tooling.

Problem: Community products need structure, identity, and flow, not just posts.

Solution: Built a platform for community connection and organized digital interaction.

Impact: Creates a product surface for community-led growth.

Problem: Generic LLM demos are not enough for real user workflows.

Solution: Built a retrieval-aware assistant architecture for context-rich responses.

Impact: Moves LLM use from novelty to product utility.