Projects
Agentic engineering in high-risk, high-security environments. Ontologies and knowledge graphs, multi-agent orchestration, NLP over regulatory and contract text, and the cloud infrastructure underneath it, across industries and company maturity levels.
What follows is the work itself, with the architecture described rather than asserted. Where a number is published it came from a repository or an engagement. Where one is missing, it is missing on purpose.
Capabilities
Ontologies & knowledge graphs
RDF graphs with SHACL validation, entailment and provenance on every fact.
Agentic orchestration
Dependency graphs, approval gates and reproducible runs across agent fleets.
NLP & information extraction
Typed extraction from documents, receipts and contracts, with spans kept.
Data science & evaluation
Eval harnesses, gold sets and published baselines rather than vibes.
Cloud infrastructure
Local-first where it matters, Fargate and Terraform where it scales.
Regulated & high-security environments
Audit trails, human gates and controls that hold up under review.
Architecture & developer experience
API design and platform engineering that other engineers can pick up.
Platforms
Systems other things are built on. Two are open source, two run internally.
Rooben
LiveOpen-source governance for AI agents in production.
A control plane between your agents and your budget. Rooben caps spend per run, standardizes how work is dispatched, and keeps an auditable record of what each agent did.
- Agentic orchestration
- Regulated & high-security environments
- Architecture & developer experience
Atlas
DeployedA knowledge bank, not an execution engine.
A local-first, RDF-grounded memory substrate. Atlas ingests documents, mail, calendar and code, resolves them into typed entities with provenance on every fact, and serves the graph over HTTP and MCP to any agent framework.
- Ontologies & knowledge graphs
- NLP & information extraction
- Data science & evaluation
- Cloud infrastructure
Keel
LiveAutonomous coding that lands as pull requests, never as merges.
A single-operator autonomous-coding and gated-deploy system. A Temporal orchestrator drains a cross-repo plan graph, executes each phase with a coding agent in an ephemeral worker, and lands the work as a pull request. Infrastructure changes pass a human-approved airlock.
- Agentic orchestration
- Cloud infrastructure
- Architecture & developer experience
- Data science & evaluation
Proof Point Studio
InternalReal software in motion, not slides in a video.
Internal tooling, not a product. An agent reads a repository’s own documentation, proposes value-ranked demo capabilities, builds each into a real interactive page, records it, and assembles the footage into a short reel.
- Agentic orchestration
- Architecture & developer experience
- Data science & evaluation
Products
Things people use. Two are live and billable, one is an open-source library, one publishes this site.
Clausebook
LiveHIPAA redlining playbooks the lawyer owns.
Build a Business Associate Agreement redlining playbook through an adaptive interview, then apply it to a real uploaded BAA. Clausebook splits the document into clauses, classifies each against an ontology and your playbook, and drafts redlines cited to HIPAA authority.
- Ontologies & knowledge graphs
- NLP & information extraction
- Regulated & high-security environments
- Data science & evaluation
Tough Crowd
LiveFeedback from 20 users, before you have 20 users.
Drop in a URL. Synthetic personas research your market, then actually use your product in live browser sessions, click by click, and return a scorecard, friction findings pinned to the step that produced them, and a ranked fix-list.
- Agentic orchestration
- NLP & information extraction
- Data science & evaluation
- Cloud infrastructure
voice-of-agents
AlphaBeta users lie. Simulate them honestly.
An MIT-licensed Python library that simulates user research with Claude, then hands the resulting personas to a live browser-based eval harness that navigates your real product. Published on PyPI.
- Agentic orchestration
- Data science & evaluation
- NLP & information extraction
Content Engine
LiveThe publishing pipeline behind this site.
A multi-tenant content service: keyword and SERP research, model-assisted drafting, Postgres storage, and a versioned read API that consuming sites fetch. This site is one tenant of it, and every article on /writing is served by it.
- NLP & information extraction
- Data science & evaluation
- Cloud infrastructure
- Architecture & developer experience
Engagements
Client work, from a consumer marketplace at scale to federal compliance decisioning. Cornerstone Mortgage and CMS are end clients reached through Policy Edge AI.
Whop
CompleteAI infrastructure for a $1.6B social commerce platform.
Executive advisory and hands-on engineering through a period of rapid scaling. Built the content moderation platform, the fraud detection system, and the foundational data platform, and led the data strategy that aligned business decisions around it.
- Data science & evaluation
- NLP & information extraction
- Cloud infrastructure
- Architecture & developer experience
Policy Edge AI
LiveCompliance AI architecture for a govtech platform.
Turned scattered regulatory documentation into an AI architecture and an execution plan, then built targeted prototypes that retired the critical technical unknowns early. The platform work reaches two regulated end clients: Cornerstone Mortgage and the Centers for Medicare & Medicaid Services.
- Regulated & high-security environments
- NLP & information extraction
- Ontologies & knowledge graphs
- Data science & evaluation
Cornerstone Mortgage
LiveCompliance decisioning for a mortgage lender.
An end client reached through Policy Edge AI, where the compliance-decisioning platform is applied to mortgage lending.
- Regulated & high-security environments
- NLP & information extraction
Centers for Medicare & Medicaid Services
LiveCompliance decisioning in a federal healthcare context.
An end client reached through Policy Edge AI. The delivery chain runs through that platform rather than as a direct federal engagement.
- Regulated & high-security environments
- NLP & information extraction
- Ontologies & knowledge graphs
“Blake came into a messy situation and quickly turned a big vision and scattered documentation into a platform we could actually deliver. By building the right prototypes early, he helped us answer the biggest open questions and move forward with confidence. He brought clarity and momentum exactly when we needed it.”
Christina Heintze Seavers, Founder & CEO, Policy Edge AI
Explorations
Ideas tested or still being tested. Some become products, most do not. Listed because the failures inform the consulting as much as the wins do.
LEAN Startup Vetting Platform
ExploringValidate demand before you build, programmatically.
Automates the fake-door test: spin up landing pages, run targeted campaigns, track real conversion signals, and find cohorts with actual purchase intent before writing product code.
Employment Law Triage
ExploringHelp individuals find out whether they have a case.
An intake system that asks the right questions and gives preliminary guidance to someone who does not know whether they have a claim. A different problem from redlining contracts for counsel, which is what Clausebook does.
Goals & Purpose SaaS
ExploringAccountability that sticks, with meaning built in.
Connects daily actions to a stated purpose, on the theory that checking boxes fails because it never explains why the box mattered.
AI in Education Launch Program
ExploringMake AI adoption safe and practical for schools.
Ready-to-deploy pilots with guardrails, plus curriculum scaffolds, for administrators who need to move without breaking trust.
Inside Politics Investment Thesis
ArchivedPolitical intelligence as structured data for investors.
Structured political data primitives for quantitative research. Archived after validating the pipeline; the insight carried into consulting work instead.
Want to dig into any of this?
The API surface for Atlas and Keel is documented, the open-source work is on GitHub, and the architecture write-ups above are the real thing rather than a summary. If something here is close to a problem you have, say so.