About the job
We're scaling the AI science behind QuickBooks' financial intelligence platform, and we're looking for a Senior Staff AI Scientist to build the probabilistic forecasting models that power cash flow prediction and financial forecasting for millions of small businesses.
Being part of QuickBooks Fintech means you'll sit at the center of decisions that directly shape small business financial health. You'll work closely with data pipeline and infrastructure teams to take models from concept to trained, calibrated, production-ready systems.
Responsibilities
Design, implement, and train large-scale forecasting models end-to-end in code.
Practices leadership and communication skills to influence teams and to evangelize AI science across the organization.
Collaborates with stakeholders to define success criteria and align model metrics with business goals.
Leads technical work of a scrum team: initiating and designing model solutions, driving end-to-end architecture designs of the team’s work, and holding the team accountable for high quality code, git, design, costs and implementation standards.
Performs hands-on data analysis and modeling with large data sets, including discovering data sources, getting data access, cleaning up data, and making them “model-ready”.
Applies data mining, NLP, and machine learning (such as supervised/unsupervised, Causal-ML, Online Learning, Bayesian Learning, Reinforcement Learning, or Deep Learning) to real-world problems and datasets.
Mentor other scientists and engineers, contributing to a high-performance, inclusive AI science community that promotes innovation and craft excellence.
Communicate AI tradeoffs and strategy clearly to technical and non-technical leaders.
Qualifications
Minimum
8+ years of experience in applied ML/AI, with a strong track record of delivering scalable, production-grade AI systems.
BS, MS or PhD in Statistics, Mathematics, Computer Science, Economics, Operations Research, or equivalent.
Deep hands-on experience building and training time-series or sequence models for multi-horizon probabilistic forecasting (e.g., transformer-based forecasters such as TFT, DeepAR, or N-BEATS/N-HiTS, or comparable architectures).
Strong applied grounding in probabilistic forecasting: quantile or distributional loss functions, calibration evaluation, and multi-task objectives across several targets.
Experience building large-scale learned representations (e.g., entity embeddings at millions-of-entities scale), including cold-start strategies.
Strong fundamentals in a modern deep learning framework, with experience building production training and inference pipelines.
Experience optimizing models for production latency and throughput constraints, from real-time inference to large-batch scoring.
Excellent communication skills and the ability to work effectively with both technical and non-technical partners, including risk and compliance stakeholders.
Preferred
No preferred qualifications listed.