Our Work
Projects, opinions, and POCs.
What we've built, what we've learned, and what we think about the technology landscape in healthcare, pharma, and manufacturing.
March 2026
Momentum, Engineering Discipline, Scale: The AI-Assisted Delivery Curve
A small AI-assisted team can deliver in a month what takes a traditional organization twenty months. That compression creates predictable dynamics across three phases, each with distinct constraints and stakeholder risks.
March 2026
The Formula: Domain Expertise + Engineering Rigor + AI
We combine domain expertise, engineering rigor, and AI acceleration to deliver production systems that rapidly create real customer value.
March 2026
Why AI Coding Fails
Because teams generate AI code without the right engineering process. We found success when we start with deepe analysis and clean documentation of existing systems, product requirement documents, clean and simple maintainable code patterns, near-100% test coverage, fully automated CI/CD pipelines, and other such rigor. This is still a 10x lift. A meaningfully net positive 10x+ lift.
March 2026
Modernizing Decades of Legacy Systems at a Large Supply Chain Company
How we are using domain knowledge and Claude Code to map, document, and begin modernizing a complex legacy landscape spanning mainframes, Lotus Notes, Azure, and offline systems.
March 2026
From Empty Repository to Production in Three Months at a Pharmaceutical Manufacturer
How domain knowledge, engineering experience, and Anthropic's Claude Code compressed a regulated software deployment from the typical twelve months to under three. Delivered as Custom SaaS, not a project.
February 2026
Most AI Failures Aren't Model Failures. They're Data Shape Failures.
Traditional data systems were built for humans. AI agents need a machine-facing layer designed for reasoning and action.
February 2026
Text, Audio, Art, Metadata: We Built an Autonomous Media Pipeline in 8 Weeks
An autonomous pipeline that runs end-to-end but knows when to ask for help. Multi-modal AI orchestration, five constraints, and what it takes to ship.
February 2026
Single Error Boundary: A Radical Approach to Debugging AI Systems
One catch block in the entire codebase. Everything else raises. Why this is the most productive debugging approach for AI systems in active development.
January 2026
Designing Zero-Egress Research Enclaves for Healthcare PHI
How we architect secure Azure environments where sensitive health data stays protected while enabling AI/ML research at scale. Architecture beats policy.
January 2026
AI Coding Tools Are Breaking Things Faster Than They Build Them
Over 90% of developers use AI tools. But CodeRabbit's research found AI-generated code creates 1.7x more issues. Here's how to unlock AI's potential safely.
December 2025
From Sketch to Launch in a Month: CPA Workflow Automation
Document AI, multi-agent review, automated client communications, and enterprise security — built for a CPA firm in four weeks.
December 2025
Why We Build Before We Pitch
The case for showing up with something wrong instead of nothing at all. How building prototypes changes the conversation.
February 2026
From Knowledge Graph to AI Tutor: Building MathPractice.ai
Custom algorithms, real-time proficiency tracking, and AI-generated practice at scale. How we built adaptive math education from first principles.
February 2026
A Layered Architecture for Self-Service Cloud Provisioning
How to build a service catalog that scales without breaking. Four distinct layers with clear responsibilities and failure boundaries.
January 2026
The Knowledge Base Pattern: How We Made AI Systems Genre-Aware Without Changing Code
Externalize domain knowledge to markdown files. Adding a new genre means adding a file. No code changes, no deployments, no prompt surgery.
January 2026
A Cloud Tagging Strategy That Actually Works
How to design, enforce, and maintain tags that don't decay into chaos. Six mandatory tags, four enforcement layers, controlled vocabularies.
December 2025
The IaC Testing Pyramid: Catching Infrastructure Bugs Before Production
A practical framework for testing Terraform and Bicep at every level. Static analysis, plan validation, deployment tests, and policy checks.
December 2025
The Algorithm Behind Adaptive Learning
Most 'adaptive' platforms just adjust speed. Real adaptation requires modeling what students know, what they don't, and why. Here's how we built it.
November 2025
Pre-Deployment Cost Estimation: Stopping Budget Surprises Before They Start
How to catch cost overruns before a single resource is created. Build cost estimation into your provisioning pipeline with Infracost.
November 2025
Digitize Without Disrupting Compliance
How we built inventory visibility for a pharma manufacturer in 10 weeks — without replacing their GMP paper records. Layer, don't replace.
November 2025
Simulating Battery Physics in the Browser
How I built an interactive BESS simulator to understand an industry I knew nothing about. Real physics, real learning.
November 2025
Dashboard Design Pattern: Progressive Disclosure and Drill-Down
How to build operational dashboards that answer questions at every level. Progressive disclosure beats information overload.
