Applied Scientist - ML/Ai

Opendoor•Toronto, ON

About The Position

We’re looking for an Applied Scientist (ALL LEVELS) to push the boundaries of applied machine learning and AI at Opendoor. While this role will have a significant impact on our valuation systems — ensuring we provide the most accurate and transparent pricing possible — the scope goes well beyond pricing. You’ll work across a range of challenging ML problems, from multi-modal modeling to operational optimization, helping us rethink how we use structured and unstructured data to make better decisions for our customers.

Requirements

Strong software engineering and coding skills in Python, with experience contributing to production codebases
Experience developing and deploying ML models end-to-end — from research and prototyping to implementation in production systems
Hands-on experience with deep learning architectures, including ConvNets, Transformers, or similar
Advanced degree (MS or PhD) in computer science, statistics, mathematics, or a related quantitative field
Solid foundation in statistics and experimental design
Strong communication and collaboration skills — you’re comfortable working with cross-functional stakeholders and can communicate technical ideas clearly

Nice To Haves

Familiarity with Pyspark and distributed data processing
Background in search, recommendation systems, or personalization
Experience working with large language models (LLMs) or vision-language models (VLMs)
A genuine interest in real estate — no prior experience required, but you'll engage deeply with housing data

Responsibilities

Design and deploy architectural improvements to our deep neural network (DNN)-based home valuation models
Build interpretable ML models that can help us explain pricing decisions to customers
Incorporate unstructured data — like images, videos, or text — into our forecasting and valuation pipelines using cutting-edge AI models (LLMs, VLMs, etc.)
Collaborate with Engineering and Ops to enhance our human-in-the-loop pricing systems
Improve the feature engineering and model training pipelines that power our production systems
Rethink our risk and optimization models using real-world data and domain insight