Head of Autonomous Carrier Operations – San Francisco, CA
Our client solves this in minutes with an army of AI agents that find nearby empty trucks, call/email them in parallel, negotiate rates, and find the best-priced truck available.
Proven Traction: Their AI platform is in production, executing thousands of live negotiations and load bookings daily for logistics companies that collectively manage over $10 billion in annual freight.
They’re backed by General Catalyst, Jawed Karim (Co-Founder, YouTube), Y Combinator and are scaling fast across the U.S. logistics network.
Head of Autonomous Carrier Operations
Job Type: Full-time; On-site
The best AI carrier sales floor in the world doesn't exist yet. You're going to train it.
Training agents will be one of the largest job categories in five years, and it's already happening. The best AI labs are spending billions on it, because the bottleneck in AI is no longer compute, it's human expertise. The models are good enough now that the only thing they're missing is the judgment the best people in a field carry in their heads.Freight is one of the biggest jobs left to do this for. Anyone can blast the same rate to 200 carriers. That's not carrier sales, and you know it. The good ones win before the first call, knowing which four to call instead of two hundred, and when to hold instead of paying up.
Only a few hundred people in the country are truly elite at this. We want to partner with the best of them and turn the way they think into the way millions of loads get covered across the freight market.
Where we are
Our AI books freight today, live and in production for some of the largest logistics companies in the US, including multiple Top 100 brokers.On our internal lane benchmark it already gets at least one vetted bid on 60 to 80% of loads, depending on the lane. We're hiring the person who gets that to 90%+ across the board.
It calls carriers, negotiates, and covers loads on its own. It can work a lane, reach carriers at the right time, and price against the market. It's good. The job is to make it as good as the best rep you've ever worked with, then let it loose to book millions of loads, like hundreds of thousands of your best reps working at once.
The real work: teaching the agent to read a load
The hard part is already built. The agent is live, moving tens of thousands of loads for some of the biggest names in freight, and it holds its own against most reps on calls and emails. What we're missing is someone genuinely elite at this to push it further and get it covering just about any load, 90%+ of the time.That's where you come in. You know the things that don't fit in a training manual, like how to read a lane, where capacity is going before the market reacts, and whether a guy actually has a truck or the rate that looks too good is about to fall apart. You work with the agent until it makes those calls the way you would.
It's the same setup the best AI labs use to get their models better. You decide how it plays each lane, watch how it does against real freight, and keep pushing until it's as sharp as you are. It gets measured against the actual market, and the standard it's chasing is you.
You'll have your own team for this, people and AI agents that pull the calls, tag what happened, and run the tests. They handle the legwork so you can focus on the judgment calls only you can make.
How we measure it
We track how the agent is doing with a lane benchmark. Instead of guessing whether it's improving, we run it on a fixed set of real lanes and check whether it actually lands a truck.These are the kinds of lanes that fill a broker's day, a mix of equipment, markets, and lead times before pickup. On each one we look at the usual numbers like coverage, time to cover, buy vs. market, and fall-off. The main one is how often the agent gets at least one vetted bid on a load.
That's running 60 to 80% right now, depending on the lane, and we want it above 90% everywhere. Whenever you change how the agent sources, times, or prices a lane, you'll see in the benchmark within a few days whether it helped.
That's the job. Get that number up, and keep it there. Everything you teach the agent is locked in for good, working every load, every day, forever.
Why it's worth your time
At a brokerage, everything you know only helps one company. You're limited by how many calls your team can make in a day, and none of it travels past your own four walls.We make the tool brokers use to cover their freight, so what you teach the agent isn't stuck at one desk. It's working millions of loads a week for every broker on the platform.
And it keeps running when you're not. You put the thinking in once and it works every load, every day, without you on the phone.
This is where the industry is going. You'd be the one who builds how it gets there.
You'll be a great fit if
You've run carrier sales or capacity at a real brokerage, and you can look at a load, know how you'd cover it, and explain why. The explanation is the actual bar, and it's harder than it sounds.Team
We're a very lean team. That includes operators who built carrier sales floors at Echo, Ryder, Transfix, and Dardini, with advisors from the founding team of Uber Freight. On the technical side, we've built AI systems for Fortune 500s at Booz & Co. and done ML research at Caltech, Rice, Waterloo, and Monash.
We're backed by General Catalyst, Y Combinator, and angels like Jawed Karim (founder of YouTube).
Compensation Details: This position offers a competitive base salary plus eqity.