AI/ML Engineer (Founding)
About the Company:
Lightning Rod Labs develops AI systems designed to predict real-world outcomes. The company trains large language models (LLMs) using scalable self-play and unstructured data, helping enterprises build custom prediction models without requiring traditional data extraction. Its approach has demonstrated that smaller models, including 14B and 32B parameter models, can outperform frontier models on prediction-market benchmarks using generated data.
About the Role:
Lightning Rod Labs is looking for professionals to build advanced AI systems that improve real-world reasoning and prediction at scale. The role focuses on developing prediction-agent infrastructure, scaling distributed model training, transforming unstructured data into training signals, and improving model accuracy through reinforcement learning and self-play. The selected candidate will take ownership of projects from data collection and experimentation through deployment.
Key Responsibilities:
Build and evolve infrastructure for training and serving prediction agents.
Orchestrate distributed training and inference across large, multi-node GPU clusters.
Transform unstructured data streams into training signals suitable for model learning and evaluation.
Design new training objectives, reward functions, and calibration methods to help models learn causal factors and generate well-calibrated probabilities.
Optimize training frameworks, fine-tune models, and evaluate performance against benchmarks.
Develop and deliver solutions across the full workflow, from data collection to deployment.
Contribute to the development of scalable systems for AI-driven real-world prediction.
Required Qualifications:
- Strong Python programming skills and experience fine-tuning large language models (LLMs).Proven track record of delivering impactful technical systems.
- Ownership mindset and ability to work autonomously.
- Demonstrated technical skills through GitHub projects, portfolios, research papers, or demos.
- Experience working beyond small-scale environments, including millions of samples and large distributed GPU clusters.
- Ability to move between rapid analysis in notebooks and building robust, scalable data and training pipelines.
- Experience with tools and frameworks such as Verl, Hugging Face, PyTorch, pandas, and standard training and data tools.
- Strong understanding of statistics, probability, data quality, and machine learning best practices.
- Familiarity with cloud platforms, including Modal, Google Cloud Platform (GCP), Microsoft Azure, and Amazon Web Services (AWS).What the Company Offers
- Opportunity to work on advanced AI training methods using reinforcement learning and self-play.
- Involvement in developing frameworks for training LLMs to make predictions from unstructured data.
- Opportunity to contribute to critical applications in areas such as defense, healthcare, and major institutions.
- Ownership of significant parts of the technology stack, from design through deployment.
- A high-trust, fast-paced work environment with autonomy to set direction and deliver work.
- Close collaboration with a small team and across product, engineering, and customer teams.
- Opportunity to take projects from research through real-world implementation.
- Location and Work Arrangement
Location: New York, NY, United States.
Work Arrangement:
Remote-first. The company has a growing hub in Brooklyn/New York City, which is preferred. Meaningful overlap with Eastern Time working hours is required.