The posting
On behalf of Behavioral AI Solutions Client #1 , SD Solutions is looking for a talented Senior Data & AI Scientist to take ownership of two of our core behavioral-modeling efforts end to end: a user-conditioned Trader World Model that learns a high-fidelity latent representation of each trader and reproduces their multi-horizon behavioral distributions, and a large-scale sequence forecaster that predicts each user's next-horizon trading activity from their intraday behavior.
SD Solutions is a staffing company operating globally. Contact us to get more details about the benefits we offer.
This is a hands-on modeling role. You will own the full lifecycle—data, representation learning, training, large-scale experimentation, and rigorous evaluation—for transformer-based models trained on hundreds of millions of behavioral samples on GPU’s.
We are specifically looking for candidates with practical experience training deep sequence and foundation models. This includes designing training pipelines, running large-scale experiments, modeling full behavioral distributions (not just point estimates), and understanding the real challenges of modern transformer training.
This is not an AI integration or prompt engineering role. We are looking for engineers and scientists who have built, trained, and evaluated models themselves, and who hold their results to a high evidentiary bar.
Responsibilities:
- Train, fine-tune, and evaluate transformer-based sequence models for behavioral prediction on large-scale trading data.
- Design feature representations and embedding strategies for structured, event-sequence behavioral signals.
- Model full multi-horizon behavioral distributions (activity, volume, PnL, exposure, hold time) rather than single point or label predictions.
- Build and beat a rigorous baseline ladder; design leakage-safe, chronologically split evaluation that earns each claim of improvement.
- Optimize large-scale GPU training pipelines (throughput, memory, dataloading, checkpoint/resume, cost) on Vertex AI.
- Collaborate closely with psychologists, bioinformaticians, AI engineers, and researchers to translate theory into production-grade AI systems.
Requirements:
Experience
- 6+ years of experience in Data Science, Machine Learning, or AI.
- Hands-on experience training deep sequence models (Transformers/attention) and/or fine-tuning foundation models.
- Strong experience with PyTorch and modern deep learning workflows (Lightning a plus).
- Experience with distributional and representation learning—modeling latent structure and full distributions, not just labels.
- Experience designing, running, and analyzing large-scale ML experiments.
- Demonstrated ownership of the complete modeling lifecycle, from data preparation through training, evaluation, and iteration.
Technical Skills
- Expert-level Python.
- Strong understanding of deep learning, representation learning, and statistical modeling.
- Experience with high-dimensional and event-sequence / temporal datasets.
- Ability to design rigorous evaluation methodologies—chronological/held-out splits, leakage control, purge gaps, and honest baseline comparison.
- Comfort with cloud GPU training and its real-world failure modes (throughput bottlenecks, OOM, dataloader/prefetch, checkpoint/resume, cost control).
Advantages:
- Zero-inflated / heavy-tailed / hurdle modeling and calibration for sparse targets.
- World-model or model-based / latent-transition (JEPA-style) modeling.
- Experience with financial, trading, or high-frequency behavioral data.
- Background in behavioral science, psychology, economics, or cognitive modeling.
- Experience with distributed GPU training (FSDP, DeepSpeed, Accelerate, DDP, etc.) and optimizing large-scale training workloads.
- Experience with LoRA, QLoRA, PEFT, RLHF, or other foundation model adaptation techniques.
- Publications or open-source contributions related to sequence models or foundation models.
Who This Role Fits:
- Someone who genuinely enjoys training large AI models—not just integrating AI APIs.
- A researcher or engineer who likes experimenting with model architectures, training strategies, and evaluation methodologies.
- Someone who thinks in terms of representations, distributions, latent structure, and generalization.
- Someone who holds a high evidentiary bar—prefers a model that provably beats a strong baseline over an impressive-sounding architecture.
- A hands-on builder who enjoys working at the intersection of research and production.
- Someone excited to own a hard, open-ended behavioral-modeling problem from the ground up.
About the company:
We are is an innovative Behavioral AI Solutions company building the next generation of AI systems designed to understand and predict human behavior. NDA, Behavioral AI Solutions, develops proprietary Large Behavioral Models (LBMs) that go beyond traditional language models by predicting human decisions and actions across complex real-world environments. Its technology combines behavioral science, psychology, and advanced machine learning to create sophisticated predictive models. The team works across the full AI development lifecycle, including large-scale data curation, representation learning, model training, fine-tuning, evaluation, and deployment.
NDA, Behavioral AI Solutions focuses on building transformer-based foundation models trained on massive behavioral datasets using advanced GPU infrastructure. Its solutions are designed to support high-impact applications across finance, defense, enterprise decision support, and other industries. The company brings together interdisciplinary experts to transform cutting-edge research into scalable, production-ready AI systems.
By applying for this position, you agree to the terms outlined in our Privacy Policy. Please take a moment to review our Privacy Policy https://sd-solutions.breezy.hr/privacy-notice, and make sure you understand its contents. If you have any questions or concerns regarding our Privacy Policy, please feel free to contact us.
