Original job description
Who we are
Kero builds real-time sports betting technology for sportsbooks around the world.
Our systems ingest live play-by-play sports data and turn it into in-game betting markets that update continuously throughout a match or game. We focus on micro markets, contextual markets, player-driven markets, and other live betting products that make watching sports more interactive, immediate, and engaging.
At our core, we are building a live betting and trading engine: real-time data, probabilistic models, automation, pricing logic, market selection, risk feedback, and product decisions all working together under production constraints.
We are a lean, distributed team operating in a fast-moving category. The right person will have a large impact.
The role
We are looking for a senior or lead-level data scientist to own meaningful parts of our sports modeling stack.
This is not a “build a notebook and hand it off” role. It is an end-to-end ownership role across research, model design, data quality, feature logic, calibration, market behavior, production handoff, post-launch measurement, and iteration.
You will work on live betting models across sports such as football, basketball, baseball, soccer, hockey, and others. You do not need to be an expert in every sport on day one, but you need to have strong sports intuition and the humility and curiosity to build deep domain understanding quickly.
The best person for this role will combine:
strong applied modeling ability;
real sports knowledge;
first-principles thinking;
skepticism around data quality;
comfort with production systems;
commercial judgment;
and the ability to use AI as leverage rather than as a substitute for thinking.
AI has changed the baseline level of technical execution. What matters more than ever is knowing what to ask, what to test, what to distrust, what tradeoffs matter, and how to turn ambiguous sports and betting problems into models that actually work in the real world.
What you will do
You will own research and modeling for live sports betting markets, from first concept through production measurement.
This includes designing targets, features, model structures, evaluation frameworks, calibration approaches, and handoff specifications for markets that may price individual plays, drives, player outcomes, game segments, full-game results, or other live betting events.
You will be expected to understand the full path from raw sports data to a live market with odds is shown to a bettor. That means caring about whether the data is correct, whether labels are clean, whether the model is calibrated, whether edge cases are handled, whether production is using the same logic as research, and whether the market performs commercially after launch.
You will work closely with engineering, trading, product, and leadership to decide what should be modeled, what should be hardcoded, what should live in the engine, what should live in the data layer, and what should not be built at all.
What makes someone great here
You think from first principles - When something looks wrong, you do not just patch the symptom. You ask why it happened, whether the same issue exists elsewhere, and whether the architecture itself is encouraging bad outcomes.
You understand that aggregate metrics can lie - A better log-loss number does not matter if it does not map to the bets we actually offer, the users who actually bet, the markets that actually settle, and the prices we actually publish.
You have a trader’s mindset - You understand that a model is not an academic artifact. It exists to price risk, support betting markets, drive engagement, and protect profitability. You care about calibration, coverage, limits, timing, availability, and edge cases because those are the things that matter in production.
You know sports - You understand why game state matters. You understand why possession, clock, field position, foul trouble, pitching changes, injuries, substitutions, depth charts, score effects, and player roles can materially change probabilities. You do not need to know every sport equally, but you need to know at least one team sport deeply and be able to acquire domain depth quickly.
You communicate clearly - You can explain a complex modeling decision to engineers, traders, product people, and non-technical leadership. You can say what is known, what is uncertain, what changed, and what decision needs to be made.
What we are not looking for
We are not looking for someone who only wants clean datasets and well-scoped tickets.
We are not looking for someone who thinks a model is done because the notebook runs locally.
We are not looking for someone who throws complex ML at a problem before understanding the sport, the market, the data, and the commercial objective.
What you should have
5+ years of applied data science, machine learning, quantitative modeling, or similar experience.
Strong Python experience, especially with pandas, NumPy, scikit-learn, XGBoost, or similar tools.
Strong grounding in probability, statistics, regression, classification, calibration, simulation, and model evaluation.
Experience working with messy real-world data where correctness, edge cases, and data quality matter.
Ability to read and reason through production code, even if you are not the primary production engineer.
Experience building models that affect real business outcomes, ideally in one or more of sports betting, trading, market making, gaming, ad tech, real-time recommendations, fintech, or similarly dynamic environments.
Strong documentation habits.
Excellent written communication.
Genuine knowledge of and passion for sports.
Nice to have
Experience with sports betting, odds modeling, trading, market making, or sportsbook products.
Experience with live sports data, play-by-play data, player props, in-game pricing, or simulation-based modeling.
Experience with real-time systems or production ML workflows.
Experience leading other data scientists, engineers, contractors, or AI agents.
Experience designing evaluation frameworks where standard ML metrics are not enough.
Experience with MLOps, model serving, data pipelines, versioning, and reproducibility.
Benefits
Fully remote work.
Flexible working hours.
High ownership and direct impact.
Opportunity to work at the intersection of sports, betting, AI, modeling, and real-time systems.
Self-improvement and learning budget.
Performance incentives.
Competitive compensation based on level and experience.
We value ownership, clarity, direct communication, and intellectual honesty. We would rather have someone surface a hard truth early than preserve a bad assumption because it is convenient.
We are a small team, which means there is nowhere to hide - but also very little bureaucracy. If you are good, thoughtful, and proactive, you will have real influence over the product and the company.
We care more about how you think than where you went to school. If you believe you can outperform someone with more traditional credentials, we are open to hearing from you.
Kero welcomes applicants of all educational backgrounds, gender identities and expressions, sexual orientations, religions, ethnicities, ages, socioeconomic backgrounds, disability statuses, and veteran statuses.
