David Martin
I build data pipelines and data products, and the applications that put them to use. A background in backend and platform engineering shapes how I build both.
I'm a data engineer who builds reliable pipelines and the systems around them: ingestion, modeling, and the applications that make data usable. Before that, I spent years in customer-facing infrastructure work, experience that now shapes how I think about failure modes and operating systems once they're live.
As a CSM, I partnered with customers on deployment issues: failed builds, routing, DNS, environment variables, edge configuration, and production incidents.
Experience tracing failures across DNS, routing, build output, auth flows, and third-party integrations when systems behave differently in prod than locally.
I design pipelines so breakage surfaces early: validation at the right boundaries, clear failure signals, and lineage you can actually follow.
Idempotent transforms, explicit grain and keys, and clear contracts between layers, so a dataset's meaning doesn't quietly drift as it moves downstream.
I started a YouTube channel to share the real engineering behind my projects: Docker ops, Astro builds, observability, and the mistakes that don't make it into blog posts.
Data pipelines, backend services, and observability tooling: systems that move and operationalize data, built to hold up under real operational load.
Observability dashboard for infrastructure health, uptime, and operational metrics, built with Next.js, PostgreSQL, Prisma, and TypeScript. What started as a quick dashboard became a lesson in modeling and aggregating fast-growing time-series data so queries stay fast as volume grows.
AWS-powered image optimization and transformation pipeline using CloudFront, Lambda, Sharp, Terraform, and S3. Rendorix actively powers the image delivery pipeline for this site’s homepage and blog assets, handling production image transformation, caching, and optimization at the edge.
Bird species classifier trained on CUB-200-2011 with EfficientNet and ResNet50. SvelteKit frontend and FastAPI inference API for top-5 species predictions.
A small HTTP caching edge in front of a GitHub API origin, with a React UI for public profile lookup. Experiments with cache hits, TTL, and stale-while-refresh using in-memory edge cache.
I run a personal VPS as a production-style lab: Apache in front, Docker Compose behind it, and mostly TypeScript services I deploy and maintain myself.
VPS dashboard
I keep a simple index at djm-apps.com so visitors can open what is running on the server without guessing hostnames.
Case studies and deep dives on data pipelines, SQL, scaling, and the operational lessons that don't show up in tutorials.
Cursor scaffolded Sentinel's observability dashboard in minutes, but the AI-generated code couldn't keep up once PostgreSQL started accumulating 20,000 health-check rows a day. This is the story of turning a working v1 into something that could actually scale.
Read now
When my dad asked me to build a site where he could publish Bible studies and schedule content himself, the hard part wasn't the design. It was building something that would keep working without me.
Read now
LLMs are great at plausible solutions, but engineers still own assumptions, trade-offs, and failure modes.
Read now