Senior Software Engineer - Predictions

The Prediction, Optimization, and Planning (POP) team builds Afresh's core replenishment technology. We built the Prediction Engine library to house the shared components that power every new solution we build: a performant data API, configurable featurization, robust and speedy forecasting, highly parallel optimization, and scalable backtesting over tens of thousands of time series.

As our product suite and customer base grow, so does the scale and complexity of what this library needs to model. It needs to gracefully accommodate predictions and simulations at different time scales (hours, days, weeks), complex data hierarchies (pallets on a truck, shelves of mangos in a store, chunks of fruit in a bowl), and endless configuration (average shelf fullness, backroom loads, truck capacities).

About the role

As a Senior Software Engineer - Prediction Engine on the POP team, you will take this core library to its next level of performance, reliability, and scalability. Your work will soon help power replenishment decisions on more than 10% of all produce sold in the United States.

What you will do:

• In your first 3 months, you might deliver a project that helps generalize model configuration, enables no-code model deploys, or vastly improves our integration testing.
• By the end of your first 6 months, you will have owned the design and implementation of large improvements and additions to the Prediction Engine. This might be an extension to our backtesting library that allows the team to easily trigger full-scale analyses over millions of store-items on a parallel cluster, or it might be an automatic data quality monitoring module that requests human input to resolve suspicious customer data.
• By the end of your first year, you will have redesigned and extended large parts of the Prediction Engine. Some example contributions include a stress-testing framework that automatically generates difficult time series and backtests our models against them, or a generalized forecaster data model that easily accommodates warehouse, bulk, and prepared items.

Skills and experience

• BS in Computer Science or relevant field
• 4+ years of professional software development experience with a proven track record of shipping high quality applications and services.
• Experience working collaboratively with machine learning engineers, data scientists, or applied scientists on large-scale software projects involving machine learning models
• Technical leadership experience and a demonstrated ability to mentor junior engineers
• Deep expertise in library design, API design, data structures, and algorithms.
• Strong familiarity with Python

Tech Stack: Our backend is pure Python (NumPy, Pandas, Torch, PySpark, Cython, orchestrated in Airflow). We use Snowflake as our data warehouse. We’d like you to have very good familiarity with Python, but many of our problems are stack-agnostic.