// Projects
Open work in governed agentic AI.
Flagship project
ARES — Agentic Reasoning & Experiential Storage
A governed semantic memory framework for AI agents. Four typed memory networks organized for per-network governance — different retention windows, different PII handling, different access policies for facts versus experiences versus beliefs. The governance stack covers the policies enterprises actually need: PII detection and redaction, GDPR/HIPAA/SOX retention, complete audit logging, token budgeting, and cryptographic right-to-forget. ARES 2.0 adds a compiled execution layer that makes governance structurally impossible to bypass.
ARES v1
The original pure-Python implementation. Four typed networks, tiered LanceDB cache, complete governance stack, and framework integration shims for LangGraph, Pydantic-AI, NVIDIA NIM, and DataRobot — each under 50 lines.
github.com/skippythemagnificent/ares →ares-memory-mcp
MCP server that gives Claude Code persistent cross-session memory backed by local LanceDB. FAILED/SOLVED/DECIDED/CONTEXT categories, semantic search, session start/end hooks. The server running in this site's own development sessions.
github.com/skippythemagnificent/ares-memory-mcp →Curriculum & enablement
NVIDIA Partner Lab Series
Seven progressively harder labs for NVIDIA partners and startup engineering teams. Each opens with a named production failure — RAG quality collapse, NIM on air-gapped infrastructure, multi-agent non-determinism in production — then builds toward the NVIDIA-stack concept that explains and fixes it. Every fix is quantified: RAGAS delta, latency Δ, rank Δ.
github.com/skippythemagnificent/Nvidia-Partner-Labs →NAT Enablement Lab Series
Three-part hands-on series for building production agentic AI with NVIDIA NeMo Agent Toolkit on DataRobot. Constructs a multi-agent Technical Proposal Generator — a declarative YAML-defined agent pattern that field engineers can demo in customer conversations and customers can replicate for their own workflows.
github.com/datarobot/enablement-NAT-afcomponents →Agent Build Clinic
Six sequential notebooks teaching production AI agent construction with DataRobot. Uses Pydantic AI and MCP. Bridges generative AI with predictive analytics — forecasting models as callable agent tools, deployable to DataRobot's serverless prediction environment.
github.com/datarobot-community/agent-build-clinic →Agentic Enablement
Foundation curriculum teaching CrewAI agent patterns — agents, tools, tasks, crews — to data scientists working within the DataRobot platform. Covers autonomous model monitoring, deployment automation, and deploying agents to production DataRobot environments.
github.com/datarobot-community/agentic-enablement →