Neil AI
Project Overview
As organizations across every industry move toward AI-first operations, most struggle with the same core challenges: keeping AI systems running reliably over time, integrating them with real infrastructure, managing computational costs, and retaining knowledge across sessions.
The Neil AI project addresses these challenges in a working lab environment. Neil is a persistent AI system that operates continuously across SEAL Lab, integrating real-time data from over ten infrastructure sources — GPU compute, network cameras, printers, knowledge bases, and cloud APIs — into a unified monitoring and automation platform.
Unlike conventional chatbot deployments, Neil manages its own computational resource consumption through a biologically inspired regulation system modeled after pinniped neuroscience. A token metabolism tracks API costs in real time and adjusts system behavior across four operational modes, from exploratory to resource-conserving, based on current workload and budget constraints.
A cognitive memory architecture with episodic, semantic, and procedural stores enables the system to retain and consolidate knowledge autonomously across sessions, eliminating the cold-start problem that limits most AI deployments.
Long-Term Goals
- Develop a generalizable framework for persistent, cost-aware AI systems that organizations can adapt to their own infrastructure
- Build autonomous memory consolidation pipelines that reduce knowledge loss between sessions
- Integrate multi-modal sensor fusion for real-time infrastructure monitoring and anomaly detection
- Reduce non-research overhead for lab members through automated documentation, scheduling, and coordination
Skills You Will Develop
- Rust systems programming and real-time terminal UI development
- LLM API integration, prompt engineering, and cost optimization
- Real-time sensor data acquisition and multi-source fusion
- Asynchronous systems design and background task orchestration
Quick Points
- Neil operates continuously, integrating live data from GPU utilization, network cameras, cloud APIs, and over ten other infrastructure sources.
- A token metabolism with hormonal analogs — adrenaline, cortisol, and allostatic load — regulates computational spending in real time, automatically shifting between exploratory and resource-conserving modes.
- The cognitive memory system uses SQLite with a knowledge graph, supporting episodic storage, fact extraction, and autonomous consolidation — solving the cold-start problem that limits most AI deployments.