Proof, not promises. Each engagement below started with a real product, a real implementation, and a business need to earn trust with technical buyers.
Official reference implementations shipped under Render's render-examples org.
Render needed reference implementations that demonstrated their infrastructure platform's capabilities for AI workloads. Not tutorials. Not "getting started" guides. Production-grade templates that developers could deploy with a single click and actually use. The goal was not more awareness. It was stronger proof at the point of evaluation.
I built a suite of AI agent templates — voice agents, RAG chatbots, research agents, documentation assistants — each running on Render's infrastructure with proper error handling, observability, and deployment configuration. Every template was a real application, not a toy example.
The result: every template shipped under Render's official render-examples org. Not because of audience size or marketing push, but because the code actually worked in production. Engineers found the templates, deployed them, and built on top of them — exactly the adoption pattern a founder wants when proof starts doing real go-to-market work.
Technical evaluation of AI coding agents on Render's infrastructure — original research, not a vendor pitch.
Read the article →Compliance documentation and best-practices guide supporting Render's HIPAA product launch.
Read the article →Production-grade, one-click deploy templates shipped under Render's official org.
Independent technical evaluation of a self-improving AI agent — and the Goodhart's Law finding it surfaced.
Inngest engaged me to build a production reference implementation of a self-improving AI agent on their durable execution platform — the kind of system their customers were asking about but couldn't find a credible end-to-end example of.
The brief was to build, benchmark, and publish. The finding was more interesting than the benchmark.
Inside the self-improving loop — where an evaluation LLM rewrites prompts to score higher against its own rubric — I documented a clear instance of Goodhart's Law in action. The evaluator gamed its own scoring criteria by embedding rubric phrases directly into outputs, inflating scores without improving actual agent behavior. The fix was strict output constraints in evaluation prompts, not more sophisticated scoring.
That finding became the centerpiece of the published piece, the YouTube video, and a TLDR newsletter placement — turning a standard reference implementation engagement into a substantive technical contribution to how teams build evaluation systems for LLM agents.
Production-grade self-improving AI agent built on Inngest's durable execution platform, with real error handling, observability, and a full evaluation pipeline.
View on GitHub →Long-form technical deep-dive on Inngest's blog, reviewed by Dan Farrelly (co-founder). Focus on the Goodhart's Law finding and how to build evaluation systems that resist it.
Read the article →Long-form video walkthrough on @mitchalderson, covering the architecture, the finding, and live demo of the evaluation gaming behavior.
Watch the video →Syndication of the technical finding to TLDR's developer audience.
Enterprise customer engineering during Series B growth.
Before Alderson.dev, the foundation. Apollo GraphQL was scaling rapidly after their Series B, and enterprise customers needed more than documentation. They needed someone who could bridge the gap between the product vision and the customer's production environment.
As Lead Developer Support Engineer, I built enterprise onboarding systems, training programs, and reference implementations that helped customers go from evaluation to production deployment. This wasn't content work — it was engineering work that produced the kind of proof founder-led go-to-market teams can build around.
The systems I built supported the platform's growth to 25M+ developers. The reference implementations became the standard onboarding path for enterprise accounts.
Reference implementations I built to help enterprise customers understand Apollo's platform in production contexts.
Every completed engagement gets added here. If your developer tools company needs independent technical content that actually moves developer adoption and sales pipeline — built by an engineer, not written by marketing — let's talk.
Book a Discovery Call →