Member of Technical Staff (Foundation Models)

Abel AI; Northern (KY); Hybrid; USD 150,000 - 230,000; Full time

Job summary

Abel AI is seeking researchers to build temporal foundation models and AI systems that understand world dynamics, forecast future outcomes, and support better decisions. You will work on scale to 10T+ observations and multimodal temporal modeling.

You will contribute to agentic forecasting frameworks, scalable inference pipelines, and research-to-system translation across modeling, data, and production. The role suits candidates with PhD or strong ML research background, experienced in PyTorch

Qualifications

• PhD or strong research background in ML, time-series, or related fields.
• ,

Responsibilities

• Scale temporal foundation models to 10T+ observations.
• Build multimodal temporal models.
• Develop agentic forecasting systems.
• Develop scalable inference infrastructure.
• Translate research into real systems.

Skills

PyTorch

scikit-learn

Python

Time-series data

Forecasting

Causal learning

Agentic AI

LLMs

Education

PhD in Computer Science or related field

Master’s candidates considered

Tools

PyTorch

scikit-learn

Python

Job description

Build foundation models and AI systems that understand how the world evolves, forecast what comes next, and support better decisions about the future.

Our mission

The world is always changing. The key is not only to observe what happens, but to understand the dynamics behind those observations, predict how they will evolve, and decide what to do next.

We aim to build temporal superintelligence: AI systems that learn from the world’s observations, understand temporal dynamics, forecast future outcomes, and reason about decisions under uncertainty.

Our goal is to push AI beyond static pattern recognition toward systems that can answer: What is happening? Why is it happening? What will happen next? And what should we do about it?

What you’ll work on

• Scale temporal foundation models to 10T+ observations. Design and train large-scale foundation models over massive, heterogeneous time-series and structured datasets, pushing scaling laws for temporal intelligence.
• Build multimodal temporal models. Develop models that jointly reason over text, time series, tabular data, events, and other structured signals for forecasting, interpretation, and decision‑making.
• Build agentic forecasting systems. Develop agentic frameworks that decompose general forecasting questions, retrieve and analyze relevant information, invoke forecasting models and tools, reason over uncertainty, and synthesize answers.
• Develop scalable inference infrastructure. Build efficient training and inference pipelines that enable low-latency, high-throughput forecasting across large numbers of models, datasets, horizons, and users.
• Advance models from prediction toward reasoning and decision‑making. Develop methods that connect forecasting with causal learning, counterfactual reasoning, uncertainty estimation, and intervention analysis.
• Adapt and specialize foundation models. Develop pre-training, post‑training, fine‑tuning, and adaptation techniques that allow general temporal models to perform strongly across new domains, datasets, and forecasting tasks.
• Translate research into real systems. Work across modeling, data, infrastructure, and product to turn research ideas into reliable systems operating on real‑world data.

What we look for

• Currently pursuing or holding a PhD in Computer Science, Applied Mathematics, Statistics, Electrical Engineering, Machine Learning, or a related field. Exceptional Master’s candidates with strong research or engineering backgrounds will also be considered.
• Deep understanding of modern machine learning, with hands‑on experience using frameworks such as PyTorch and scikit‑learn.
• Strong engineering fundamentals and excellent Python expertise, with the ability to build research systems that scale beyond prototypes.
• Experience working with time‑series, tabular, or other structured data, including forecasting, representation learning, or large‑scale data modeling.
• Experience with LLM or foundation-model pre‑training and/or post‑training, including large‑scale training, fine‑tuning, alignment, or model adaptation.
• Understanding of causal learning, including causal discovery, causal inference, counterfactual reasoning, or related methods.
• Hands‑on experience building agentic AI systems, including tool use, retrieval, planning, reasoning workflows, or multi‑agent architectures.
• Ability to work across the full ML stack—from data and model architecture to distributed training, evaluation, inference, and deployment.
• Strong research intuition combined with the ability to ship working systems.
• Experience training large‑scale foundation models from scratch.
• Experience with distributed training, large‑scale data pipelines, GPU clusters, and model optimization.
• Research experience in time‑series foundation models, forecasting, multimodal learning, causal learning, or generative modeling.
• Experience with efficient inference techniques such as batching, compilation, quantization, distributed inference, or model serving.
• Experience building agents that combine LLMs, forecasting models, code execution, retrieval, and external tools.
• Publications at top‑tier machine learning venues such as NeurIPS, ICML, or ICLR, or significant contributions to widely used open‑source ML systems.

The problems we care about

Today’s foundation models are remarkably good at understanding what has already been written. We want to build models that understand what happens next.

That requires learning from observations across time, discovering the dynamics that generate them, reasoning about uncertainty and interventions, and turning those predictions into decisions.

Understand the past. Model the dynamics. Predict the future. Decide what comes next.