Business intelligence · Regulated deep-tech
Director of Business Intelligence by title; in practice, a systems-level decision architect for regulated deep-tech. Under the BI umbrella here: production AI and agentic workflows, scenario and TCO engines, compliance evidence, and analytics where the numbers must hold. One thesis — align data, evidence, rules, and economics before capital commits.
Shipped: 151+ tests on an agentic CRM (KineticCRM) with multi-strategy recovery, 105+ dbt quality gates on EcoMetrics, national-scale logistics and EPR compliance on a RENW-class platform, plus public testimony and primary-source privacy/AI-governance research. Real systems, not slides.
BI here is not "another dashboard." It is the system through which complex decisions get made under regulation, scale, and capital constraints — comparable scenarios, traceable economics, evidence order, and kill criteria.Four capability lanes carry the work: commercial wedge, decision infrastructure, regulatory strategy (including compliance engineering, AI governance, and privacy), and builder systems (production AI, agentic workflows, analytics).
Production AI and agentic workflows with real recovery paths and serious test coverage (151+ tests on KineticCRM). TCO and scenario engines, analytics with data quality gates (105+ dbt checks on EcoMetrics), logistics and EPR automation, and the regulatory and commercial story so the numbers survive scrutiny.Expect contracts, tests, evals, and observability. Not one-off demos.
I work full time in industry. This site is mostly documentation of how I build. If you have a concrete problem in climate or materials, enterprise ops, compliance-heavy product, or an AI/ML platform that needs senior production rigor, send a note. I reply async.
Founders, operators, and diligence teams in materials, packaging, and bio-industrial usually start with one of four capability lanes. Pick the lane that matches your problem, then follow the evidence on this site.
Regulated domains where bench, data, rulemaking, and go-to-market cannot drift apart. Each lane is a way I work, not a job title.
Build the first real product-market argument for feedstock × product × buyer.
Structured pathways, assumptions, evidence order, and kill criteria.
Rules as competitive infrastructure — across materials regulation, privacy, and AI governance.
Production AI, agentic workflows, and analytics where the numbers and behaviors must hold.
Tier 1 work spans climate analytics, national logistics and EPR automation, and CRM automation with tests behind it. Try EcoMetrics live; project pages include Executive vs Engineer detail where it helps.
Climate intelligence: validate environmental claims for 75+ SKUs. dbt pipeline with 105+ data quality tests, DuckDB analytics, Streamlit dashboards — transaction-level logic, monthly aggregation models, multi-company architecture.
Logistics and operations at enterprise scale: unified shipping engine for 1K–100M units, piecewise volume discounts, multi-modal transport, geospatial port detection (Haversine), EPR integration, 75+ SKU catalogue — production BI with Prometheus/Grafana.
Agentic AI systems: CRM where autonomous agents manage leads and workflows — multi-strategy error recovery (~85%), graph-based workflow execution, conversation analysis, and 151+ automated tests.
Research-side projects (trading agents, spatial P2P, vision pipelines, portable GIS) and in-development theses (Providence, Craton). Most feed lessons back into production work; none of them belong on the front page.
Selective conversations with founders, operators, and diligence teams in materials, packaging, and bio-industrial. Brittle workflow, fuzzy economics, regulatory load, or a pathway that needs a kill criteria? Mention your lane on the contact page. I reply when there is a real problem on the table.
Prefer async? Use email on the contact page. A calendar slot is there if a live conversation helps.