Senior AI Engineer
PITZ | Remote | Reports to CTO
A trillion-dollar industry still runs on phone calls. We're building the intelligence that replaces it.
PITZ is the infrastructure layer for the global automotive aftermarket. A mechanic speaks — in Spanish, Portuguese, or English, in a noisy shop — and our system diagnoses the failure, identifies the exact part, selects the supplier, underwrites credit, and executes the order. No screen. No typing. No human in the loop.
We're live in Mexico and Brazil. Entering the US this year. And we need someone to build the brain behind all of it.
Why this role exists
We're not hiring a cog for an AI team. There is no AI team yet. You'd be the first dedicated AI hire. You'll work directly with the CEO to design and build the entire intelligence layer — voice pipeline, recommendation engine, credit scoring, and the data flywheel that makes everything compound.
If you deliver, you become Head of AI within 12 months. Not as a title bump — because you'll have built the thing from scratch and no one will know it better than you.
This is a role for someone who's been a Senior ML Engineer or AI Lead somewhere and knows they're ready for more — but hasn't had the chance to prove it because there's always someone above them.
Here, there's no one above you. Just the problem and the product.
What you'll build
The voice-to-action pipeline. Mechanic speech in noisy shops → intent extraction → diagnostic prediction → SKU mapping from millions of parts → order execution. Three languages. Regional slang. Background noise. Real world, not a lab.
The recommendation engine. Right part, right seller, right price, right delivery. This is where margin lives. Every percentage point of improvement moves real revenue.
The credit scoring models. Behavioral underwriting for mechanics with no credit history. Turn operational data — order frequency, payment patterns, voice engagement — into financial signals.
The data flywheel. Every interaction makes the system smarter. You design the architecture that turns daily transactions into compounding intelligence.
US market integrations. Connect PITZ to Identifix, ALLDATA, OBD platforms, PartsTech, Mitchell1. Adapt models for the US vehicle fleet and English-speaking mechanics.
What you bring
3-5 years building ML/AI systems in production. Not notebooks. Not demos. Systems that real users interact with daily and that break in interesting ways.
NLP or voice experience. You've worked with speech recognition, intent extraction, text classification, or conversational AI. Bonus if it was in noisy, multilingual, or domain-specific environments.
Full-stack ML instincts. You can train a model, build the data pipeline, optimize inference, deploy it, and monitor it. You don't wait for a platform team — you are the platform team.
Builder mentality. You've worked in startups or fast-moving environments. You ship fast. You make decisions with 70% of the data. You'd rather build something imperfect today than plan something perfect for next quarter.
Fluent English required. Spanish or Portuguese strongly preferred. You'll work with mechanics, ops teams, and partners across three countries.
Requirements
Timezone: Must be available during US/LatAm business hours (UTC-3 to UTC-8). This is a real-time collaboration role, not async.
Commitment: This is a full-time, dedicated position. If you're currently juggling multiple roles — this isn't a fit. We need 100%.
Travel: Periodic travel to Mexico City and São Paulo. You'll visit workshops, hear mechanics, and see the product in action. If you've never left your desk to watch a user interact with your model — this will change that.
What makes this different
At most companies, a Senior AI Engineer optimizes a model inside a product someone else designed. Here, you design the product. The intelligence layer IS the product. There's no PM handing you a spec. There's no VP of AI telling you the roadmap. You figure out what to build, you build it, and you see it run a real supply chain the next day.
You'll work with proprietary data that doesn't exist anywhere else — millions of real mechanic-to-system interactions mapped to parts, orders, payments, and outcomes. Your models get better not because of more compute, but because of more reality.
And if you prove you can own this — you lead the team we build around it.