Software engineer focused on distributed systems, cloud infrastructure, and AI.
Neural Network Observability (paper) AI models can detect their own mistakes internally, but whether that signal is accessible depends on architecture choices made before deployment. Some architectures make it readable, others don’t. Current interpretability methods can’t recover a signal that the architecture didn’t produce, making neural network training design an AI safety decision. Four model families, 11 scales. Built with Python, PyTorch.
NxtChess (demo) Multiplayer chess platform with custom AI playstyles, anonymous play via shareable links, and OAuth 2.0 auth. Built with SolidJS, Go, PostgreSQL, Redis, and WebSockets.
Compass Autonomous agent architecture for real-time decision-making in a 3D MMORPG. Layered decision stack with GOAP planning, Monte Carlo robustness gating, Bayesian online learning, and JPS/A* pathfinding. Built with pure Python 3.14, zero external dependencies.
Material Type Prediction with WiFi/BLE Early wireless sensing experiment using commodity WiFi/BLE RSSI to detect objects and infer material class.
Carmichael, T. "Architecture Predicts Linear Readability of Decision Quality in Transformers." Preprint, 2026. [Zenodo] [PDF] [Code]
Dai, G., Paluri, P., Carmichael, T., Cheng, A., Miikkulainen, R. "Leveraging the Selfless Driving Model to Reduce Vehicular Network Congestion." IEEE Real-Time Systems Symposium (RTSS), 2019.
Email: hello@tdosmail.com