Principal AI Researcher (Post-Training Foundation Models)

Stabile Search•New York, NY

5d•Onsite

About The Position

My client is one of the most prestigious and reputable quantitative trading firms in the world and they are at the forefront of AI and Machine Learning investment and development. They are looking for an experienced ML Research Engineer who specializes in post-training foundation Large Language Models. If you are excited about a role where you will have ownership over groundbreaking new projects in a fast-growing team, then this is the opportunity for you. Compensation $2,000,000 - $5,000,000 Total Cash Compensation depending on level of experience. As a ML Researcher in their new and fast-growing AI division, you will be post-training foundation Large Language Models (LLMs) that will power quantitative research analyzing market data across the firm. They are investing heavily into this area and are attracting many of the world's top AI talent to their team, so you will be surrounded by great talent. They are also committed to investing heavily into the tools and infrastructure needed to be a leader in this space.

Requirements

Experience post-training models.
Architecture Expertise: Deep command of how modern LLMs work under the hood — transformer internals, attention, and the mechanics that drive model behavior — in both theory and practice.
Post-Training Experience: Direct, hands-on work adapting models after pre-training, spanning supervised fine-tuning, parameter-efficient approaches (e.g., LoRA), and preference-based alignment methods such as RLHF and DPO.
Agentic & Workflow Design: Track record building orchestration frameworks, reasoning strategies, and structured workflows that power intelligent assistants or automated analysis systems.
Systems & Hardware Fluency: Comfort operating close to the metal — managing GPU memory, working across precision formats (FP16/BF16, quantized models), and applying distributed training and parallelization strategies.
Engineering Toolkit: Expert-level Python alongside modern ML frameworks — PyTorch, JAX, the Hugging Face ecosystem, DeepSpeed, or similar.
Quantitative Foundations: Strong grounding in the mathematics and statistics underpinning quantitative finance.
Strong problem-solving skills and a results-oriented mindset.
Excellent communication skills and ability to work in a collaborative environment.
PhD or Masters in a related field.
Finance-related domain experience is not needed but interest in trading and finance is helpful.
Strong interest leveraging Machine Learning modeling to own and make a large impact in a quantitative finance setting.

Responsibilities

Model Customization & Fine-Tuning: Take their foundation LLMs and tailoring them to proprietary datasets, with the goal of achieving best-in-class results on challenging financial language and structured/tabular data.
Agentic Reasoning & Tool Use: Build orchestration layers and intelligent workflows that allow models to reason through multi-step analytical problems and effectively leverage external tools and data sources.
Training & Evaluation Infrastructure: Develop end-to-end pipelines for training and benchmarking LLMs, with particular emphasis on reasoning ability, quantitative/mathematical performance, and factual accuracy.
Model Alignment: Apply and experiment with techniques like RLHF, DPO, and self-improvement/bootstrapping methods to steer models toward finance-specific goals.
Retrieval & Knowledge Integration: Build retrieval-augmented generation (RAG) systems — including embeddings and vector stores — that connect in-house models to broader knowledge sources.
Productionization: Drive down inference latency and optimize model performance under demanding real-world constraints, partnering with engineering teams to deploy solutions into live trading and research workflows.