The posting
Principal AI Engineer
Job Summary
As a Principal AI Engineer, you will be the most senior individual contributor in AI engineering, responsible for shaping the applied research direction for Task-Specialized Models (TSMs) and the ML systems powering the platform. You will drive innovation in model distillation, inference economics, reinforcement learning, and system deployment, directly impacting customer metrics and enterprise solutions. This role combines hands-on research, system design, and cross-functional collaboration to translate cutting-edge AI research into scalable, real-world applications.
Responsibilities
- Own the SMA Research Agenda: Define and execute the research roadmap for Specialized Model Projects, focusing on domain distillation and accuracy-per-dollar improvements tailored to enterprise needs.
- Push Distillation and Model Compression: Design and implement teacher-student distillation programs, optimize models through quantization, pruning, and other compression techniques, and demonstrate their effectiveness on production workloads.
- Advance Reinforcement Learning and Preference Optimization: Develop RL and preference learning pipelines (e.g., RLHF, DPO), manage reward modeling, training loops, and evaluation methodologies to improve agent learning from deployment feedback.
- Own Inference Economics: Lead efforts to reduce cost-per-inference, latency, and increase throughput through model and serving-side optimizations, including hardware selection and system-level improvements.
- Bridge Research and Production: Convert research prototypes into production-ready models, collaborating with model runtime teams and deployment pods to ensure customer metrics improve.
- Set the Technical Bar: Lead design reviews, experiment evaluations, and technical decision-making to uphold high standards in applied ML.
- Grow the Team: Mentor senior and research engineers, contribute to hiring, and foster a culture of rigorous experimentation and applied impact.
- Represent the company Externally: Publish research, present at conferences, and engage with the AI research community to enhance the company’s technical reputation.
Qualifications
- PhD in an AI-Adjacent Field: Machine learning, deep learning, NLP, large language models, reinforcement learning, knowledge distillation, model compression, ML systems, multimodal learning, agent-based AI, or information retrieval. Prefer candidates from top institutions such as MIT, Stanford, UC Berkeley, CMU, Princeton, Harvard, Oxford, Cambridge, ETH Zurich, etc.
- Applied Research Track Record: 8+ years of combined PhD and industry experience or 5+ years post-PhD focused on shipping ML systems. Publications at top venues (NeurIPS, ICML, ICLR, CVPR, ACL, EMNLP, AAAI) are highly valued.
- Production ML Experience: Proven experience deploying ML models in production environments, including fine-tuning, distillation, and serving at scale.
- Programming & ML Frameworks: Expert in Python, with deep familiarity with PyTorch, HuggingFace Transformers, DeepSpeed, vLLM, and related tooling. Ability to read CUDA and Triton kernels or collaborate with kernel specialists.
- Distillation & Model Efficiency: Hands-on experience with knowledge distillation, quantization, pruning, and inference optimization stacks like TensorRT, ONNX, or TGI.
- Reinforcement Learning Expertise: Practical knowledge of RL algorithms (PPO, DPO, RLHF, RLAIF), with experience extending or implementing these methods.
- Evaluation & System Design: Skilled in designing meaningful evaluation metrics and system architectures for large-scale distributed training.
- Communication Skills: Strong written and verbal communication, capable of producing technical papers, design docs, and leading technical reviews.
Preferred Skills
- Prior applied research or engineering experience at leading AI labs or AI-first product companies.
- Contributions to open-source ML projects.
- Experience collaborating with academia on publications or standards.
- Startup or entrepreneurial experience with shipping products under ambiguity.
Experience
- 8+ years of combined PhD and industry experience, or 5+ years of post-PhD applied experience in shipping ML systems.
- Demonstrated success in deploying models at scale, especially in enterprise or large-scale environments.
- Proven track record of research impact through publications, citations, or open-source contributions.
GrowthOpportunities
- Opportunities to lead cutting-edge research initiatives.
- Potential to influence product strategy and enterprise AI solutions.
- Pathway to senior leadership roles within the AI organization.
