Applied AI/ML Scientist, Intern

Faire•San Francisco, CA

4d•Hybrid

About The Position

Our Applied AI/ML Science team builds and maintains the models that power the marketplace. That work includes the shipping and delivery estimates retailers rely on as they shop and check out: what it will cost to ship and when it will arrive, before the order has been packed. Within Applied Science, our Shipping and Fulfillment team builds the models behind those estimates, including models that learn to understand products from images and text. Better predictions give retailers confidence in what they're buying and, in turn, help brands sell more on Faire.

Requirements

Currently enrolled in or recently graduated from a Master's or PhD program in Computer Science, Machine Learning, Statistics, Electrical Engineering, or a related technical field
Hands-on experience building deep learning models (e.g., PyTorch), ideally with images, text, or both, including fine-tuning pre-trained models or working with embeddings
Familiarity with gradient-boosted trees and other tabular ML methods
Strong Python and SQL
Ability to read research papers and turn promising ideas into working code
Solid grounding in statistics and model evaluation, including benchmarking against strong baselines and estimating uncertainty
Comfort working with noisy or incomplete real-world data
A genuine enthusiasm for tackling ambiguous problems and learning new tools and techniques

Responsibilities

Predicting how an order will be packed: Before an order ships, we have to anticipate how it will be packed. You'll build models that learn from what each item is and how items combine in a cart.
Recommending better ways to pack: How an order is packed shapes what it costs to ship. You'll develop models that understand items from images and text and learn from historical packing outcomes to recommend packing that reduces shipping cost and improves efficiency.
Understanding products from a catalog: Listings often lack reliable weight, size and shape. You'll train multimodal deep learning models that infer these physical characteristics from images, text and other catalog signals.
Survey the literature and existing approaches to identify promising ideas
Prototype and train models offline, benchmarking against our current methods
Build out the strongest approach into a working implementation
Work with Applied Scientists and ML Engineers to test it against live traffic
Present findings and recommendations to the team