Machine Learning Engineer

We are on a mission to create the forecasting foundation model to rule them all. Forecasting drives critical decisions worldwide - impacting staffing, supply chain management, finance and more. Our solution provides companies with the models, platform and APIs they need to easily generate the most accurate forecasts possible, helping to significantly reduce waste and enabling smarter, more confident decisions.

The forecasting model is at the heart of our technology. As the second founding MLE, you will build, train and deploy large foundation model architectures: implement and combine ideas from the literature, push the state of the art, and ultimately deploy your model for our customers to use in production. Our goal is for our models to be the best for our customers’ use cases - including for capabilities that do not exist yet in academic models.

You love your craft, have high standards, stay up-to-date with the latest ideas in ML, and know when to make trade-offs to ship. You live and breathe neural networks, and speak PyTorch or Jax. You are used to diving deep in large amounts of data, and you know what you train your models on. Bonus if you have experience building solid ML infrastructure.

You are passionate about your craft, maintain high standards, stay current with the latest tech and know when to make trade-offs to deliver results efficiently. We do not believe great engineers are “jack of all trades”, but rather that they excel at diving deep into complex topics quickly, leveraging a broad range of experiences to solve challenging problems. You are also open to exploring new concepts, technologies, and enjoy quickly throwing prototypes together to kick the tires. You prefer quick feedback loops, rather than aiming for perfection on the first try.

• Architect and train time-series foundation models using diverse datasets, integrating multimodal inputs like numerical time series, text, location and image data

• Stay up-to-date on the foundation model literature

• Design reproducible experiments to verify, compare and combine ideas from the literature.

• Build your own data exploration tools to understand (lagged) correlations between different data sources, data sparsity, weather patterns, consumer trends…

• Add data sources you find interesting to our train or test datasets.

• Deploy models for use in our API and platform - getting into the gritty details if exporting to ONNX requires some custom operation or torch.compile fails

• Gather and act on user feedback, iterating on model capabilities to maximize customer satisfaction and impact.

• Mentor and guide future team members, helping shape a high-performing science and engineering culture as the team grows.

• You have extensive ML engineering and science experience.

• You have worked on training large foundation models.

• You are intimately familiar with Python, PyTorch or Jax.

• Strong drive and ambition - you really care about having an impact on the product, and revenue.

• You go deep. You enjoy understanding all the details, and you don’t make assumptions without checking them.

• You are intimately familiar with at least one low-level language such as CUDA, Rust or C++.

• You’ve worked extensively with time series data and systems delivering real-time data streams.

• You have started projects from scratch. You might have led a project, been a founder previously, or built an impressive side project with real users.

• You have built RAG-ready models.

• Be part of an elite, diverse and fun team. We celebrate having people with different backgrounds who are the best at what they do. We especially enjoy working with cracked people with nonlinear paths. Our CTO was a theater consultant in another life. 🎭We already hail from 6 nationalities with only 5 people. Come be our 7th!

• Make the world more efficient. In business, success is driven by smart decision-making. By providing the most accurate insights as early as possible, we help ensure that more decisions lead to successful outcomes, creating a positive impact worldwide.

• Satisfy your curiosity. Learn how the world works, from within household-name companies.

• Build state-of-the-art systems. Shape our infrastructure to handle and deliver state-of-the-art multi-modal forecasting systems.

• Competitive compensation

• Generous equity

• Daily lunch vouchers

• Gym in the office

• Monthly contributions to a mobility pass

• Full health insurance for you and your family

• Tasty caffeine in the office

Our goal is to provide the most accurate and easy-to-use forecasts to our customers, by leveraging refined information from their own industry. Foundation models for time series are changing this entirely. With the current advances in data processing and model training, we can now pre-train models on diverse temporal data across industries. We provide value to our customers by enabling rapid interaction with our models when provided data and context in natural language, delivering real-time forecasts with accuracy reports. Our customers do not need to be data scientists or have a PhD in Machine Learning to build and ship an accurate forecasting system for their use-cases.

Example use cases include demand forecasting for sporting goods chains, large furniture companies, predicting sales and customers for a restaurant group and revenue forecasting in the gaming industry.

The founders Geoff and Joachim are both Machine Learning PhDs who have built forecasting and ML systems from scratch at JP Morgan, Amazon, Google, Bloomberg, and Sonos in the US.

We are a global company that happens to be HQed in Paris. Get the best of both worlds — Silicon Valley work ethic and ambition in the center of Paris, right across from the historical Stock Exchange, in the Sentier.