Every finding is testable.
- If I can't show you the data, I don't claim it
- Every recommendation ships with a reproducible check
- Before/after metrics, not before/after screenshots
I'm Jacob Carpenter — the engineer you hire, the engineer who does the work, the engineer who sends the invoice. No account managers, no delivery leads, no junior contractors billed at senior rates. Just me, deep in your stack, for as long as it takes to ship the fix.

I've spent the last six years building serverless architectures and distributed systems for startups — most of them small, a few of them venture-backed, all of them with a common problem: the gap between what the deck promised and what the system could actually do was enormous, and nobody wanted to close it.
TetsuKod.ai is me, working directly with founders and engineering leads who want that gap closed. Not a deck, not a recommendation, not a "roadmap" — the actual code, merged, with tests. I do the audit. I write the PRs. I run the deploy. Then I show you what I changed and why.
I've also spent a lot of time codifying how I work. Out of that came AIDE — a methodology and toolchain for running coding agents in a disciplined way — and aidemd, the server that ships it to other engineers' agents. Every audit I run uses it. It's not a product pitch; it's the reason I can do a two-week audit and show up with merged code at the end instead of a slide deck.
Solo consultancy. Deep technical audits, infrastructure hardening, and agentic-AI integration for teams that need senior-level work without a senior-level hire. Also building the underlying ops platform — lead scoring, outreach, DMARC/SES, ActivityPub-federated blog — that runs the business.
Refactored legacy Node monoliths into modular Lambda functions. Built a RAG chatbot on LangChain for client support. Shipped e-commerce plugins across WordPress, Shopify, and Wix. Introduced Zod-based schema validation across teams.
Ran sprints, reviewed PRs, and stitched CI/CD into Jira so deploys stopped being theatre. Sustained throughput across a distributed team while keeping production code reviewable.
Built an event-driven distributed system on RabbitMQ + Lambda, standardized a 10k-concurrent S3 ingest pipeline on SNS, added end-to-end encryption for sensitive user data, integrated Stripe subscriptions, and migrated infrastructure to self-managed servers for cost control.
Browser automation, scraping, and moderation bots for social-media operators and e-commerce pricing. Learned the expensive way how brittle systems fail in production — and how to build the kind that don't.
30 minutes, free, no sales deck. Bring a URL and we'll find three things to fix on the call.
Book strategy call →