Confido is the AI infrastructure for consumer-packaged-goods brands — the single source of truth that 200+ leading brands use to run finance, accounting, sales, and operations. We're backed by a Series A led by Footwork and Y Combinator.

The AI/ML team builds the systems that turn Confido's data into automation: structuring information, forecasting demand, and running LLM-based agents that carry out real operational work. AI here isn't a demo feature — it powers workflows our customers depend on daily.

The Technological Challenge

CPG runs on a sprawling, disconnected set of systems, and Confido sits across the whole lifecycle. That gives us an unusually broad and connected view of the data:

• Financial operations — payments, deductions, disputes, invoices
• Retail sell-through and inventory — sales and stock signals across the retail landscape, through Muffin, our analytics product for consumer brands
• Supply-chain and demand signals

Very little of it arrives clean or standardized — which is exactly what makes it interesting to model.

What you might work on

• Document understanding — turning messy real-world documents into structured, reliable data
• Forecasting — demand and sales prediction across many noisy, real-world series
• Agents — systems that reconcile, draft, and act inside the product
• Anomaly detection — surfacing out-of-stocks and other signals from retail data

We care as much about whether these systems are reliable as whether they're clever; evaluation and monitoring are part of the work, not an afterthought.

How we build

We combine frontier foundation models with our own ML models and large operational datasets, in production.

Tools: Python · Ruby · Snowflake · dbt · AWS · LangChain

🌱 Open Roles