Figures Estimates
A compensation prediction engine, built on gradient boosting.
I started in finance: four internships between 2015 and 2017, in investment banking, equity research, M&A and then venture capital at Kima Ventures. That is where I learned to read a company through its numbers first, and through its founders second.
In 2017 I joined Comet, which was part of the Kima Ventures portfolio, on the operations side, and ended up managing a team of six. That is where I moved into data: trained by the Head of Data, I learned SQL and contributed to the models of the company's first data warehouse, on BigQuery, dbt and Looker.
From 2020 to 2022 I worked freelance, between data and product. For Cigusto, I held the product role on an ERP built from scratch: the link between the retailer and the development agency, from the first prototype to rollout, including training teams all over France. For Figures, I worked on the MVP and the business plan. I also helped other brands on short assignments, in food service, clothing and hairdressing.
Since 2022 I have been a Data Analyst at Figures. As the only member of the data team, I own data transformation and reporting, from dbt models to the analyses delivered to customers, and I built several tools there that are used every day, internally and by customers.
A compensation prediction engine, built on gradient boosting.
A chatbot built on a semantic layer, which handles nearly all of the team's data requests.
The same approach, opened up to Figures users.
The product role on an ERP built from scratch for the purchasing hub and the stores: purchasing, inventory, point of sale, invoicing, promotions. Rolled out to 120 stores in two weeks.
The link between the business, the ERP and logistics during the move to a new logistics platform: order and stock flows with the ERP, cleaning the product and EAN database before launch, and planning orders ahead to cover several days of network downtime.
Metabase set up during the project: monitoring of cash and returns anomalies, and store footfall heatmaps compared with staff schedules.