Senior Data Scientist
Eat Club
Data Science
Sydney, NSW, Australia
Posted on May 15, 2026
About EatClub
At EatClub, we believe restaurants and bars are the beating heart of every city's culture. Whether it's discovering a hidden gem, grabbing a late-night takeaway, or meeting friends for a drink, our mission is simple: help the hospitality industry thrive through smart, powerful tech.
Our platform helps over 2 million customers discover top restaurants and access real-time deals that save them up to 50% off the bill. We empower more than 4,000 venues to fill empty tables, increase foot traffic, and maximise revenue.
Recently ranked #11 on the 2025 Deloitte Tech Fast 50. Now is an exciting time to join our team. Initially co-founded by Marco Pierre White and leaders in the food-tech scene, we're now a 150+ person scaleup growing fast and making waves in the industry.
Why You'll Love Working With Us
We are looking for a Senior Data Scientist with strong ML engineering skills - someone equally comfortable doing the hard forecasting and statistical work as they are shipping production infrastructure.
A Day-in-a-Life of our Senior Data Scientist
EatClub's forecasting product is at an inflection point. A TiDE-class challenger model is materially better than production but not yet shipped. The MLOps platform exists as a draft RFC, not a running system. Your job is to change that.
You'll spend most of your time making restaurant revenue forecasts more accurate, more explainable, and faster to ship. Some weeks that means deep modelling work (feature engineering, model selection, evaluation design). Other weeks it means MLOps (deployment pipelines, monitoring, the per-venue model selector). Always with a production endpoint in sight.
There's ambiguity. There's speed. There's ownership.
You won't be handed perfectly defined tickets. You'll help define them.
You will work closely with the Senior ML Engineer who owns production and daily calibration, as a peer. Your combined focus moves between getting EatLab live on AWS, resolving the TiDE-vs-LightGBM model choice per venue cohort, and building the variance attribution work that lets operators understand why a forecast missed. Occasionally, BD will pull you into a beta venue conversation. That feedback sharpens what you build next.
You'll regularly work with:
Production forecast accuracy moves into the "very good" band. Operators trust the forecast enough to act on it. Models ship from notebook to production in days. The MLOps platform exists and other teams want to use it. Variance attribution is live and the Conversational Venue Assistant can answer "why did the forecast change today" with grounded reasoning. The team is genuinely deep on ML.
Maybe this role is not for you if
One last note: even if you feel you don't meet every criterion above, we encourage you to apply. Past work experience is not the only indicator of future success, and we are on the lookout for hungry talent who want to grow with us.
At EatClub, we believe restaurants and bars are the beating heart of every city's culture. Whether it's discovering a hidden gem, grabbing a late-night takeaway, or meeting friends for a drink, our mission is simple: help the hospitality industry thrive through smart, powerful tech.
Our platform helps over 2 million customers discover top restaurants and access real-time deals that save them up to 50% off the bill. We empower more than 4,000 venues to fill empty tables, increase foot traffic, and maximise revenue.
Recently ranked #11 on the 2025 Deloitte Tech Fast 50. Now is an exciting time to join our team. Initially co-founded by Marco Pierre White and leaders in the food-tech scene, we're now a 150+ person scaleup growing fast and making waves in the industry.
Why You'll Love Working With Us
- Join a small forecasting team at the moment when accuracy lift directly converts to addressable market
- Own the ML stack end to end: modelling, MLOps platform, deployment, monitoring
- Work on a genuine industry gamechanger that "closes the loop" between demand prediction and revenue generation
- Operate in an AI-first culture (Claude Code via Anthropic Enterprise, agentic and async workflows expected)
- Ship in days, not quarters. Real customers, real feedback every week
We are looking for a Senior Data Scientist with strong ML engineering skills - someone equally comfortable doing the hard forecasting and statistical work as they are shipping production infrastructure.
A Day-in-a-Life of our Senior Data Scientist
EatClub's forecasting product is at an inflection point. A TiDE-class challenger model is materially better than production but not yet shipped. The MLOps platform exists as a draft RFC, not a running system. Your job is to change that.
You'll spend most of your time making restaurant revenue forecasts more accurate, more explainable, and faster to ship. Some weeks that means deep modelling work (feature engineering, model selection, evaluation design). Other weeks it means MLOps (deployment pipelines, monitoring, the per-venue model selector). Always with a production endpoint in sight.
There's ambiguity. There's speed. There's ownership.
You won't be handed perfectly defined tickets. You'll help define them.
You will work closely with the Senior ML Engineer who owns production and daily calibration, as a peer. Your combined focus moves between getting EatLab live on AWS, resolving the TiDE-vs-LightGBM model choice per venue cohort, and building the variance attribution work that lets operators understand why a forecast missed. Occasionally, BD will pull you into a beta venue conversation. That feedback sharpens what you build next.
You'll regularly work with:
- The CTO on architecture decisions and the MLOps platform direction
- The Senior ML Engineer on model selection, hyperparameter tuning, and feature engineering
- Backend engineers on POS data pipelines and serving APIs
- The PM on what to measure, what to ship, and what to defer
- The BD lead and real restaurant operators (occasionally), to understand where the forecast lands or misses in their day
- Run a model evaluation across hundreds of venues, decide which approach wins on which cohort, and ship the result
- Build out the internal MLOps platform: workflow orchestration, feature store, model registry, serving
- Productionise a new modelling approach (e.g. TiDE-class neural forecasters) and validate compute footprint before deploy
- Land "why we missed" variance attribution so the product can explain its own predictions
- Push the team's AI-first workflow forward: agentic loops, async runs, humans on final review
- Production demand-forecasting for thousands of venues across multiple POS systems (Square, Lightspeed, H&L, Ideal) enriched with weather, events, and holidays
- The internal MLOps platform: workflow orchestration, feature store, model registry, serving, monitoring
- Per-venue model selection: deciding which model architecture wins for which venue cohort and operationalising the choice
- Variance attribution and forecast explainability, including the LLM-and-vector-DB direction for "why did this prediction change"
- Hourly / intraday demand modelling on top of the daily forecast
- The Actions Feed intelligence layer: turning forecast deltas into ranked, executable recommendations
- Forecast confidence: quantile loss, P10 / P90 bands, calibrated per-day confidence scores
- Strong Python and modern data tooling fluency (pandas / polars, pydantic, pandera, pytest)
- Demand-forecasting depth as your primary strength: comfortable with LightGBM and gradient boosting for tabular forecasting, plus at least one of Prophet, AutoARIMA, AutoETS, or TiDE-class neural forecasters. Quantile loss, exogenous regressors, time-series cross-validation second nature
- Statistical rigour: experiment design, hypothesis testing, and causal reasoning
- Production MLOps experience: shipped models end to end, with hands-on experience in workflow orchestration, experiment tracking, and model serving. Databricks experience is a strong plus - that is where our ETL and data platform lives
- Kubernetes + Terraform in production, on AWS
- Software engineering rigour: type checking, unit tests, CI/CD, code review discipline
- Strong "bias to action" and shipping evidence (not RFCs, shipped systems)
- "AI-first" working style: Claude Code, agentic workflows, AI in your daily loop
- LLM, RAG, or vector-DB experience (we have a real use case in variance attribution and the Conversational Venue Assistant)
- Databricks exposure (ETL lives there)
- Hospitality, retail, demand-forecasting, or marketplace ML domain experience
- "E-shaped generalist" breadth: data science + ML engineering + data engineering + analytics + software engineering
- Experience setting up an ML team or pairing with an existing DS without territorial dynamics
- Defaulting to the shortest path to a measured result in production
- Comfortable working alongside an existing strong ML Engineer as a peer, not over them
- Direct, low-ego, willing to be wrong in public
- Curious about the actual problem (restaurant operators making better decisions) not just the modelling artefact
- Treating AI tools as leverage, not as a novelty
Production forecast accuracy moves into the "very good" band. Operators trust the forecast enough to act on it. Models ship from notebook to production in days. The MLOps platform exists and other teams want to use it. Variance attribution is live and the Conversational Venue Assistant can answer "why did the forecast change today" with grounded reasoning. The team is genuinely deep on ML.
Maybe this role is not for you if
- You prefer research over shipping
- You're uncomfortable owning ambiguous problems end to end
- You're uncomfortable working alongside an existing strong ML Engineer as a peer
- You've never owned an end-to-end production ML stack
- You want to focus purely on modelling or purely on infrastructure - this role requires both
One last note: even if you feel you don't meet every criterion above, we encourage you to apply. Past work experience is not the only indicator of future success, and we are on the lookout for hungry talent who want to grow with us.