Building Klego's AI application builder
My ongoing contract work across an AI application builder, including model integrations, isolated previews, version history, GitHub synchronization, and AWS delivery.
[01] $ blogs.fetch --feed=latest
[02] > opening content store.............OK
[03] > sorting by published date.........OK
[04] > computing read time...............OK
[05] > rendering post grid...............OK
[06] > feed ready........................OK
> [BLOGS] activating....
[01] $ blogs.fetch --feed=latest
[02] > opening content store.............OK
[03] > sorting by published date.........OK
[04] > computing read time...............OK
[05] > rendering post grid...............OK
[06] > feed ready........................OK
> [BLOGS] activating....
> Project stories and engineering notes
First-person project stories and notes on full-stack and Generative AI engineering. I write about the systems I build, the parts I own, and how the application fits together.
> showing 1–5 of 12 posts · page 1 of 3
How I built Raccog as a solo founder, connecting Python APIs and document-based RAG with the platform's generation, preview, storage, and publishing features.
My work delivering the Solviser React Native app and NestJS backend end to end, with the application available on Android and iOS.
Most AI support deployments plateau at 20% deflection. The teams that actually hit a 70% reduction in support effort run all four layers — deflection, triage, copilot, bounded auto-resolution — gated by real evals and confidence thresholds. The architecture, the failure modes, and the honest math.
A practical guide to running n8n locally, persisting workflows properly, and deploying production-grade instances on Docker, AWS, Railway, and Kubernetes — with queue mode, secrets, and CI for workflow versioning.
Browse my projects or get in touch about a senior engineering role or product you need help building.