AI Engineer

About the Role

The AI Engineer will be instrumental in designing and developing proprietary machine-learning models. These models will interpret electrical and mechanical engineering work, including schematics, PCB layouts, and CAD geometry, for use in automated scoring and evaluation systems.

Responsibilities

• Design and build proprietary machine-learning models capable of reasoning about electrical and mechanical engineering artifacts for automated scoring.
• Develop novel model architectures and training pipelines focused on technical reasoning, extending beyond text generation to include multi-modal understanding of CAD/ECAD data, simulation outputs, and user interaction traces.
• Translate complex engineering tasks (e.g., circuit design, debugging, system integration, mechanical design tradeoffs) into machine-interpretable representations for reliable and deterministic evaluation.
• Manage the complete learning loop for these models, encompassing data generation from assessments, trajectory capture, failure analysis, curated dataset creation with golden supervision, and continuous model evaluation and iteration.
• Create robust, defensible, and difficult-to-replicate scoring systems that form the core of our assessment platform.

Requirements

• Strong foundation in machine learning, including deep learning and modern foundation model architectures.
• Proven experience in designing and operating end-to-end training and evaluation pipelines for production ML systems.
• Practical experience with retrieval-augmented generation (RAG) systems and vector databases for large-scale knowledge and artifact retrieval.
• Experience working with noisy, real-world labeled datasets, including data cleaning, schema design, and quality control.
• Hands-on experience with reinforcement learning, specifically including one or more of: reinforcement learning from human feedback (RLHF), preference modeling and reward model training, or policy optimization for multi-step/tool-using agents.
• Experience building or training models on multimodal data (text, images, video, or structured technical artifacts like diagrams or CAD files).