Ben Battles
AI operator · instructor · consultant
About
Product leader and full stack builder who takes AI systems from prototype to production. Demonstrated experience in large-scale agent orchestration, teaching applied gen AI, and consulting on AI strategy. Years spent across LATAM leading multilingual teams, studying cultural differences, and collecting too many plants.
Built in California, building in Mexico
Work
Agentic Systems
Agent architecture for messy, high-volume operations — built to scale.
Applied AI
Teaching founders to build AI products — zero to shipped.
AI Strategy
AI tooling, vendor evaluation, model selection, and system design.
Agentic Systems
Agent architecture for messy, high-volume operations — built to scale.
Applied AI
Teaching founders to build AI products — zero to shipped.
AI Strategy
AI tooling, vendor evaluation, model selection, and system design.
Portfolio
Try my skills
Installable Claude Code skills I've built — workflow automation, code review, morning briefings, and more. Clone the repo, drop the skill in, start using it.
Browse on GitHubCacti OS
Download and set up your own AI chief of staff. Clone once, run one command, walk away with a fully wired personal AI system — knowledge base, agents, and daily briefings included.
Clone on GitHubLive Apps
Rivers of Civilization
SoonExplore how rivers shaped the ancient world.
Coming soonSystems I've Built
Production AI systems and designs I've shipped.
Scheduling Agent
Ops coordinators were manually emailing 30+ facilities a day to book pickup and delivery appointments.
Replies are sorted into confirmations, counteroffers, FCFS notices, and auto-replies — edge cases route to a human.
Live in production — schedules the majority of appointments end-to-end with no human touch.
Automated Quoting Engine
We were winning only a tiny fraction of spot freight loads on a major account — manual quoting couldn't keep up with automated bidders.
A self-improving system — it analyzes its own wins and losses, proposes new pricing rules, and tracks rule changes automatically. Built to become a second brain across the company.
Took win rate from near-zero to competitive — auto-bidding live on all spot loads for two national accounts.
Tender Intake Pipeline
Every shipment began with ~15 minutes of reading a tender PDF and hand-entering lane, facility, and cargo details — 10+ times a day.
Custom dashboards and automated processes run the pipeline end to end — extraction, matching, and shipment creation.
Live in production — auto-builds shipments straight from tender PDFs, cutting ~15 minutes of manual entry per shipment to near-zero.
Appointment Prediction
The scheduling agent could request appointments — but not decide when. A human still picked every date, blocking full automation.
Transit math is FMCSA hours-of-service compliant — the model orders stops and predicts dates before a human has to.
Live in production — closes the last human decision in the loop, enabling fully autonomous scheduling.
Agentic System Exercise
Most AI courses stop at theory — learners never ship a working agentic system of their own.
Two templates do the heavy lifting — one frontend, one backend — so learners finish with a complete agentic workflow they built themselves.
Built to teach founders at AI Study Camp — download it, follow the guide, and walk away with a working agentic system, front to back.