Hi, I'm @joseph,
AI Developer
who builds self-healing
autonomous & AI systems.
Career Highlights
Two industry-defining milestones — the start and the frontier of the self-driving era.
Present at the founding.
Part of the original Google team that hit Larry Page's 100,000-mile no-incident milestone — the result that green-lit the self-driving car project.
Guinness World Record.
Helped execute the world's first commercial delivery by a self-driving truck (Otto → Uber ATG, 2016) — a driverless Budweiser haul across Colorado.
Built the future from scratch.
Part of the team that built Google's Firefly — a ground-up, fully-electric autonomous vehicle with no steering wheel or pedals.
Coast to coast, no hands.
Original sub-10-person Pronto.ai team behind the self-driving Prius that drove San Francisco → New York with no hands on the wheel and no pedal input.
Worked alongside people who now lead Aurora, Waymo, Nuro, Kodiak, and Wayve.
Work Experience
Previously: 10 years as an autonomous vehicle systems specialist at Google, Uber, and startups — followed by 5+ years learning AI tools, code, and building personal and client projects.
Founder and Builder
2023 - Present · AI / Software
- •Operate Hermes (Nous Research's open-source agent runtime) as an AI operations team — Lurkr as CTO, C-suite skills, VPs, 40+ scheduled automations handling real business operations
- •Develop private AI solutions for SMB clients using Ollama, Open WebUI, and local automation tools
- •Ship production web and mobile apps: React, Next.js, React Native, TypeScript, Supabase
- •Created Whisper Walkie — open source push-to-talk voice typing tool (100% local processing)
- •Built The Skate Workshop — React Native app for Olympic coach Willy Santos (19-table Supabase backend, push notifications, live on iOS via TestFlight, Android builds rolling)
- •Cloud Infrastructure — Terraform multicloud (AWS/GCP), Kubernetes, Anthos/GKE
Key technologies
Hermes — Agent System
I operate Hermes as CTO. Executive skills, three VPs, 40+ scheduled automations, Ollama Cloud models pinned per job. Built on Nous Research's open-source agent. Migrated from OpenClaw in 2026.
- Daily standup and evening-plan briefings compiled from every subsystem
- Inbox and email watch with digest-first triage (draft, never auto-send)
- A steward agent per project — daily status, blockers, and escalation
- Weekly executive reviews, scorecards, and report generation
- Watchdogs monitoring the crons, uptime, and model health — alerts only when wrong
- Deployment pipelines: automated CI/CD
Private AI Solutions
Private AI solutions for SMBs. Self-hosted open source LLMs, client data never leaves their servers.
- Ollama + Open WebUI: full chat interface with local models
- Local automation workflows: AI-powered business process automation
- Local RAG: embeddings over client documents
- Integration with existing systems (CRM, email, etc.)
- Client team training for adoption
Production Applications
Complete applications built and shipped. React/React Native frontend, Supabase/PostgreSQL backend, external API integrations.
- The Skate Workshop: React Native app on a 19-table Supabase backend (live on iOS via TestFlight, Android rolling)
- DALL-E Image Generator: web app with OpenAI DALL-E 3.0 API
- Cloud Infrastructure: Terraform multicloud (AWS/GCP), K8s
- Whisper Walkie: voice-to-text 100% local (open source)
Specializing in private AI setups and multi-agent workflows for SMBs. Delivering 100% local, self-hosted LLMs that process sensitive data without ever leaving the client's servers.
Autonomous Vehicle Technician / Operations
2018 - 2019 · Autonomous Vehicles
Original sub-10-person team. Sole technician — owned end-to-end fleet integration, validation, and test campaign execution under rapid iteration.Drove the 2,900-mile San Francisco → New York autonomous demo with zero critical failures and zero hands on the wheel.Conducted 30+ executive and investor demonstrations as primary field-to-engineering liaison.
2,900-Mile Cross-Country Demo
Supported a 2,900-mile cross-country autonomous demonstration with zero critical failures. Conducted 30+ executive and investor demonstrations.
Watch Demo Video"Sole technician operating Pronto.ai's entire fleet — built and integrated the autonomous system that drove coast-to-coast with zero critical failures."
Autonomous Truck Technician / Operations Lead
2016 - 2018 · Autonomous Vehicles
Helped execute the world's first commercial delivery by a self-driving truck — a Guinness World Record driverless Budweiser haul across Colorado (Oct 2016).Led technician team installing wiring harnesses and mounting compute + sensors onto a 10-truck autonomous fleet, maintaining ~90% uptime.Installed newly-developed drive-by-wire systems and wheel-encoder infrastructure; vetted a manufacturing vendor in Juárez for scaled harness production.Primary field-to-engineering liaison: daily failure triage, root cause analysis, and resolution through the Otto → Uber acquisition.
"End-to-end ownership of a 10-truck autonomous fleet — from drive-by-wire integration to daily operations and root cause analysis."
Program Manager L4 / Operations & Sensor Readiness Lead
2009 - 2016 · Autonomous Vehicles
Owned full sensor-suite readiness — cameras, IMUs, three LiDAR variants, radar — as final gate before vehicles hit public or private roads.Built calibration infrastructure from scratch: targets/rigs from 80/20 extrusion, bench tools, and an anechoic chamber for radar.Cut full vehicle calibration from 9 hours to 60 minutes fleet-wide across 100+ vehicles; trained the operations team and ~100 drivers.Co-built the drive-by-wire system and installed onboard compute, including testing new NVIDIA chips for the self-driving stack.
"Supported the world's first fully driverless ride on public roads (2015) — a blind passenger was the first user."
Projects
joblasPrivate AI Solutions
Private AI solutions for SMBs using Ollama, Open WebUI, and local automation tools. Client data never leaves their servers.
- 100% local deployment — no cloud data leakage
- Local automation workflows
- Local RAG: embeddings over client documents
- Integration with existing client systems
- Client team training for adoption
for SAs
Hermes
AI operations system I run as CTO. Executive skills, VPs, 40+ scheduled automations, Ollama Cloud models pinned per job. Built on Nous Research's open-source Hermes agent. Migrated from OpenClaw in 2026.
The Skate Workshop
Full-stack React Native mobile app for Olympic-level skateboarding coach Willy Santos. 19-table Supabase backend, live push notifications, 4,900+ trick-obstacle training combinations, 175 automated tests. Live on iOS via TestFlight; Android builds rolling.
cv-joseph
Interactive CV with AI chatbot, agentic RAG, voice mode, Langfuse observability, and multi-layer defense. SSR prerender, bilingual i18n.
DALL-E Image Generator
AI image-generation web app using the OpenAI DALL-E 3.0 API — a demo build.
Whisper Walkie
Push-to-talk voice typing tool with 100% local processing. No data sent to the cloud.
Cloud Infrastructure
Multicloud Terraform deployment (AWS/GCP), K8s migration with Cloud SQL HA, legacy modernization with Anthos/GKE.
AI Coding Power User
High-Agency · AI-NativeAI coding assistants are my core development tools. Multi-agent orchestration to full-stack apps — AI amplifies every phase.
- ›Agent operations: one orchestrator + 40+ scheduled automations running real business operations on Ollama Cloud
- ›RAG systems: embeddings, similarity search, context-augmented generation
- ›Prompt engineering: system prompt design, evaluations, quality/cost optimization
- ›Full-stack AI apps: prototype to production using Claude, OpenAI, Ollama Cloud, and open source models
Hermes Command Center
How I run the business with AI agents
Hermes is my agent operations system. Lurkr is the main agent, with executive skills, three VPs, and 40+ scheduled automations. Each job is pinned to the Ollama Cloud model that fits it.
JTS Org Chart v5.1 — role structure across Engineering, Infrastructure, Product, Marketing, and Personal
Ollama Cloud Model Selection
Every scheduled job and the main agent are pinned to a model that fits; a deep fallback chain covers outages.
Gateway sessions and most scheduled jobs
Spawned worker sessions and coding tasks
First in the fallback chain
Resilience when the primary provider degrades
Education
10 Years AV + 5+ Years AI/Code
No formal degree. 10 years hands-on in autonomous vehicle programs (Google/Waymo, Uber, Pronto.ai) + 5+ years learning AI tools, code, and building personal and client projects.