I turn messy, cross-system workflows into applied-AI products that people can inspect and control. I own the path from workflow analysis through interfaces, integrations, tests, and operational safeguards.
Wachter Intelligence · Selected systems
Guarded agent actions. A Go CLI and portable agent skill that separates model planning from deterministic spend checks, cart writes, read-back, bounded rollback, and human-controlled checkout. The visual uses a reproducible local mock run with sanitized demo data. Source and architecture →
Long-running model workflows. A terminal transcription tool with audio preview, long-file splitting, speaker handling, rate-limit preflight, and SRT or text output. The proof card shows the real interface with a sanitized demo transcript, not a claimed model result. Source and setup →
| System layer | What I own |
|---|---|
| Agent systems | Tool boundaries, deterministic validation, read-back verification, failure stops, and human approval. |
| Operational products | Workflow discovery, interfaces, backend services, data flow, and third-party integration. |
| Long-running AI workflows | Input preparation, chunking, model calls, split/upload/transcribe progress, recovery, and structured output. |
| Delivery and reliability | Tests, health checks, staged deployment, monitoring, backups, and operating documentation. |
Selected private work includes voice and receptionist automation with telephony, booking, follow-up, persistence, monitoring, and deployment safeguards. The repositories, client details, provider topology, and operating data stay private.
TypeScript · React · Node.js · Python · FastAPI · Go · Postgres · Docker · Cloudflare · telephony and model APIs
I run Wachter Intelligence, an applied-AI systems studio focused on operational workflows. For technical review, start with Carrito’s public safety architecture or the Transcribe setup. You can also find me on LinkedIn.