Associate Applied AI - Deep Learning/Time Series

Job Description

WHAT WE DO:

At Goldman Sachs, our Engineers don't just make things – we make things possible. Change the world by connecting people and capital with ideas. Solve the most challenging and pressing engineering problems for our clients. Join our engineering teams that build massively scalable software and systems, architect low latency infrastructure solutions, proactively guard against cyber threats, and leverage machine learning alongside financial engineering to continuously turn data into action. Create new businesses, transform finance, and explore a world of opportunity at the speed of markets.

Engineering, which is comprised of our Technology Division and global strategists groups, is at the critical center of our business, and our dynamic environment requires innovative strategic thinking and immediate, real solutions. Want to push the limit of digital possibilities? Start here.

AI RESEARCH AT GOLDMAN SACHS: 

The AI Research group is the firm's dedicated research organization, operating at the intersection of frontier machine learning and quantitative finance. We build, train, and rigorously evaluate deep learning models on some of the richest financial time series data in the industry — market microstructure, cross-asset pricing, macroeconomic indicators, transaction flows, and alternative data.

Our mandate is to advance the state of the art in sequence modelling and probabilistic forecasting for noisy, non-stationary, low signal-to-noise financial data, and to deliver that research as a firmwide platform that quantitative researchers, strategists, and engineering teams across the organization can build on. We operate with research rigor and engineering discipline — every model we ship is reproducible, benchmarked against strong baselines, and evaluated under realistic out-of-sample and out-of-regime conditions.

THE ROLE:

Title: AI Research – Associate  Location: New York, NY  Division: Engineering – AI Research 

We are seeking a deeply hands-on researcher to lead the design, training, and evaluation of deep learning models for financial time series. This is an individual contributor role for someone who is equally comfortable deriving a likelihood, writing a distributed training loop across a multi-node GPU cluster, and defending an evaluation methodology to a room of quantitative researchers.

You will own research problems end to end: framing the question, curating and engineering the data, designing the model architecture, running large-scale training experiments, building the evaluation harness, and partnering with quant and engineering teams to bring models into production. Because our output serves multiple desks and asset classes, you will be expected to build models and abstractions that generalize — not one-off solutions. 

This is a fast-moving research space. You thrive in ambiguity, you are skeptical of results that look too good, and you bring the same rigor to evaluation methodology that you bring to model design.

WHAT YOU WILL BE WORKING ON: 

• Model research and development: Design, implement, and train modern deep learning architectures for forecasting, representation learning, and generative modelling of financial time series — including CNNs and temporal convolutional networks, Transformers and attention-based sequence models, autoencoders, GANs, diffusion models, graph neural networks, Bayesian networks, and reinforcement learning.

• Time series specialization: Build and benchmark against specialized sequence architectures including WaveNet, N-BEATS / N-HiTS, DeepAR, PatchTST, and Time Series Foundation Models (TSFMs), and rigorously baseline them against classical econometric methods such as ARIMA, GARCH, Kalman filters, and state space models.

• Training at scale: Own large-scale model training across the firm's GPU clusters and cloud compute environment — distributed data parallel (DDP), fully sharded data parallel (FSDP), mixed precision, hyperparameter search, and experiment tracking. Optimize inference through quantization, distillation, and ONNX-based deployment paths.

• Evaluation and validation: Build robust evaluation frameworks tailored to financial data — walk-forward and purged cross-validation, embargo periods, regime-conditional analysis, uncertainty quantification and calibration, ablations, and significance testing that properly accounts for multiple hypothesis testing and data snooping.

• Applied quantitative research: Partner with quantitative researchers and strategists across desks on alpha research, signal generation, portfolio optimization, and backtesting, translating model outputs into economically meaningful, risk-adjusted, capacity-aware signals.

• Firmwide research platform: Contribute reusable models, datasets, benchmarks, and tooling to a shared research platform that serves multiple desks and asset classes, raising the quality and reproducibility bar across the firm.

• Technical leadership: Mentor junior researchers and engineers, review research designs and code, and present findings to senior technical and business stakeholders.

• AI Governance: Ensure all models adhere to the firm's model risk management, data privacy, ethics, and safety standards, with full documentation, lineage, and auditability.

SKILLS AND EXPERIENCE WE ARE LOOKING FOR: 

Required

• A Bachelor's, Master's, or Ph.D. degree in Computer Science, Machine Learning, Statistics, Mathematics, Physics, Electrical Engineering, Quantitative Finance, or a related quantitative discipline.

• A minimum of 3+ years of industry experience building, training, and deploying deep learning models, with a substantial portion focused on sequential or time series data. Candidates with a Ph.D. and fewer years of industry experience will be considered where the depth of research experience is demonstrably equivalent.

• Deep expertise across modern deep learning architectures: CNNs and TCNs, Transformers, autoencoders, GANs, diffusion models, GNNs, Bayesian methods, and reinforcement learning.

• Strong working knowledge of classical time series and econometric modelling — ARIMA, GARCH, Kalman filtering, state space models — and clear judgment on when deep learning does and does not beat them.

• Practical experience with modern time series architectures such as WaveNet, N-BEATS / N-HiTS, DeepAR, PatchTST, or Time Series Foundation Models.

• Expert-level Python and deep proficiency in PyTorch, TensorFlow / Keras and/or JAX / Flax.

• Demonstrated experience with distributed and accelerated training (DDP, FSDP, multi-node GPU training) and model export and optimization workflows including ONNX.

• Rigorous foundations in statistics, probability, stochastic processes, optimization, and signal processing.

• Excellent oral and written communication skills, with the ability to articulate research trade-offs to both PhD researchers and non-technical business stakeholders.

• Comfort operating in a quickly evolving environment with a high degree of ambiguity and rapid change.

Preferred 

• Publications at top-tier venues such as NeurIPS, ICML, ICLR, AISTATS, or KDD, or in leading quantitative finance journals.

• Prior machine learning or data science experience at a hedge fund, asset manager, proprietary trading firm, or systematic trading desk. 

• Experience across multiple asset classes — equities, fixed income, FX, commodities, or multi-asset.

• Familiarity with Bayesian deep learning, probabilistic programming, or conformal prediction for uncertainty quantification.

• Experience with cloud platforms and containerized training environments (AWS, Kubernetes, Docker).

• Exposure to model risk management or regulatory frameworks in financial services.

• Contributions to open-source machine learning or time series libraries.

WHAT'S IN IT FOR YOU: 

• Work on genuinely hard research problems with immediate, measurable commercial impact across the firm.

• Access to proprietary, high-quality cross-asset financial datasets at a scale few institutions can offer, and GPU compute to model them properly.

• Join a standalone research group that values rigor — strong baselines, rigorous evaluation, and reproducibility.

• Freedom to publish and engage with the broader research community.

• Direct partnership with quantitative researchers and strategists across multiple desks who will use what you build.

• Develop deep expertise at the intersection of frontier deep learning and quantitative finance, with the scope to shape the firm's applied AI research agenda.

Salary Range  
The expected base salary for this New York, New York, United States-based position is $100,000-$200,0000. In addition, you may be eligible for a discretionary bonus if you are an active employee as of fiscal year-end.

ABOUT GOLDMAN SACHS:

At Goldman Sachs, we commit our people, capital and ideas to help our clients, shareholders and the communities we serve to grow. Founded in 1869, we are a leading global investment banking, securities and investment management firm. Headquartered in New York, we maintain offices around the world.

We believe who you are makes you better at what you do. We're committed to fostering and advancing diversity and inclusion in our own workplace and beyond by ensuring every individual within our firm has a number of opportunities to grow professionally and personally, from our training and development opportunities and firmwide networks to benefits, wellness and personal finance offerings and mindfulness programs. Learn more about our culture, benefits, and people at GS.com/careers.

We’re committed to finding reasonable accommodations for candidates with special needs or disabilities during our recruiting process. Learn more: https://www.goldmansachs.com/careers/footer/disability-statement.html