Blog
Field notes from shipping enterprise AI
What we're learning building agentic AI, ML, and data systems in production — written by the people doing the work.
AI-Enabled Last-Mile Manufacturing Resource Planning
ERP tells you what happened. The last mile of manufacturing planning — today's machine down, tomorrow's rush order — still runs on spreadsheets and gut feel. Here is how we combine LLMs with operations research to fix that.
June 14, 2026
HumanEval and the Vibe-Coding Gap
Models saturate coding benchmarks while AI-written codebases quietly rot. The gap between passing HumanEval and shipping production software is where the real engineering lives.
June 8, 2026
Collections AI Under Compliance Constraints
Collections is the most heavily regulated conversation in consumer finance — and one of the highest-ROI places to deploy AI. We show how to build collections automation where the compliance constraints are enforced in code, not in training decks.
June 8, 2026
StrataHub vs Building an Internal Data Team
Every AI ambition runs through data engineering. Should you build the data team first, or ship AI use cases that force the data foundation to materialize? We argue for the second — with evidence.
June 1, 2026
AI Quality Control on the Production Line
Vision-based defect detection demos well and fails in production for predictable reasons: drifting lighting, rare defects, and no plan for the false-positive bin. Here is how we build AI quality control that survives the line.
May 4, 2026
Credit Risk Models Live and Die on Data Quality
In credit risk, the model is rarely the problem. We lay out the data-quality failures that actually sink risk models in production — and the engineering discipline that prevents them.
April 27, 2026
StrataHub vs Offshore AI Development
Offshore rates look unbeatable on paper. For AI systems, the rate card is the wrong unit of analysis — here's where the economics actually land, and where offshore still makes sense.
April 13, 2026
Our Experience with On-Demand GPUs for Open-Source AI Models
A field report from six months of running open-source model workloads on rented GPUs: what it costs, where it breaks, and when it beats both the hyperscalers and buying hardware.
April 13, 2026
Making Machine Data Useful for Manufacturing AI
Factories generate terabytes of sensor and PLC data that mostly goes nowhere. The gap between 'we have the data' and 'a model can learn from it' is a data engineering problem with a known shape — here is how we close it.
March 23, 2026
From Rules Engines to ML in Fraud Detection
Ripping out the rules engine is the wrong move. We walk through the migration path that actually works: ML and rules running together, measured against each other, in production.
March 23, 2026
StrataHub vs Traditional AI Consulting Firms
Traditional consultancies sell strategy and staff-hours; we sell deployed systems. A concrete look at how the two models differ in incentives, deliverables, and what you own at the end.
February 23, 2026
AI Scheduling for High-Mix Low-Volume Manufacturing
High-mix low-volume shops are where naive scheduling AI goes to die. What actually works: constraint solvers for the core, ML for the inputs, and agents at the edges.
February 23, 2026
Intelligent Document Processing for Loan Origination
Loan files are still built by humans re-keying PDFs. We break down how to get document automation past the demo and into a production origination pipeline — with the eval harness that keeps it honest.
February 16, 2026
SKU-Level Demand Forecasting in Manufacturing
Aggregate forecasts look great in the S&OP deck and fall apart at the SKU level, where production actually gets scheduled. What it takes to forecast thousands of SKUs — including the intermittent long tail — in production.
February 9, 2026
StrataHub vs Hiring an In-House AI Team
Building an internal AI team takes 9-12 months before the first production system ships. Here's an honest comparison of when to hire, when to partner, and how to do both.
January 19, 2026
Predictive Maintenance: From Pilot to Production
Most predictive maintenance pilots die in the gap between a promising notebook and a plant floor that trusts the alerts. Here is how we close that gap.
January 19, 2026
AI Underwriting: Compliance and Explainability First
AI underwriting fails in production for regulatory reasons, not technical ones. Here is how we build models that survive both the risk committee and the examiners.
January 19, 2026