About the role
The price of physical capacity is still set by phone calls, stale rate tables, and guesswork. You'll build the index that changes that. When Juno finds open space on a truck already headed from Dallas to Houston, your models decide what that space is worth, how likely a shipper is to book it, and how likely the carrier is to accept, all within seconds. Every quote, win, and loss flows back into the index and makes it sharper.
Freight is the first market. Over time the same index will price warehouse space, production capacity, and work for autonomous systems. You'll work directly with Phillip and our engineering leads, and you'll own pricing from research through production.
Your first 90 days
Ship an improvement to the production pricing model, such as freight-class-aware pricing or better coverage on sparse short-haul lanes, measured on live win rate.
Build the evaluation and backtesting framework for the index: win rate, margin, and carrier acceptance by lane, benchmarked against market rates.
Own the next-generation index architecture, including how prices form on lanes with little history, how the model learns from every quote outcome, and how it extends to capacity beyond trucks.
What you’ll do
- Model the market. Price capacity from quotes, conversions, carrier acceptance, telemetry, and market rate data.
- Optimize the trade-off. Balance win probability and margin together, with well-calibrated uncertainty.
- Solve cold start. Price new lanes and regions with little or no history by learning across geography, freight class, and equipment.
- Close the loop. Design the experiments and feedback systems that let the index learn from every outcome.
- Ship to production. Serve prices within tight latency budgets inside live rate APIs, and monitor for drift and data failures.
- Extend the index. Generalize pricing from freight to other forms of physical capacity on IOI.
Skills you’ll need
- Probabilistic and Bayesian modeling, with a focus on calibrated uncertainty.
- Pricing, auction, or market microstructure modeling: demand curves, win probability, elasticity, and bid optimization.
- Machine learning for tabular and time-series data (gradient boosting, GLMs, neural models).
- Causal inference and experiment design, including bandits or A/B testing in live markets.
- Strong Python (NumPy, pandas or polars, scikit-learn, PyTorch) and SQL.
- Production model deployment: low-latency inference, monitoring, drift detection, and retraining pipelines.
- Geospatial and network data modeling.
Nice to have: quantitative trading or market making; marketplace pricing in rideshare, delivery, freight, or ad auctions; operations research and integer programming; reinforcement learning; Java; freight rate data (SMC3, DAT).
Who you are
- 4+ years building pricing, forecasting, or trading models that ran in production with real money on the line. Relevant PhD research counts.
- An advanced degree in a quantitative field is a plus.
- You obsess over calibration and backtests, and you know the ways a model can fool you.
- You explain model behavior clearly to engineers and to commercial teams.
- You want your models to move real goods.
Compensation
$220K to $320K plus meaningful equity. Health, dental, and vision. San Francisco, in person.
How we hire
- Meet Phillip, Oway's founder.
- Meet our engineering leads, usually one call with one or two people.
- A modeling exercise on an anonymized quote dataset, with details shared after step two.
About Oway
Oway is building the coordination OS for the physical economy. IOI connects the legacy systems industry runs on with fleets, factories and robots through one open schema, Juno is the industrial AI that acts on it, and Oway OS is where the work runs. We’re backed by Y Combinator and General Catalyst.