Relevant Experience & Education Highlights

Extensive experience building scalable AI training infrastructure, including distributed LLM training with PyTorch, GPU clusters managed by Kubernetes and Slurm, and MLOps tools, positions this candidate strongly for advancing Figure's Helix Team training systems.

1. Design, deploy, and maintain training clusters

Managed high-performance computing environments and GPU clusters optimized for deep learning workloads.

• Built next-gen AI cloud infrastructure featuring GPU clusters orchestrated with Kubernetes at a stealth AI startup.
• Developed ML serving platforms incorporating Ray, Kubernetes, and vector databases for large-scale model deployment.
• Led construction of unified control planes for machine learning platforms supporting multi-GPU distributed training at a major ride-sharing company.

2. Architect and maintain scalable deep learning frameworks for training on massive robot datasets

Architected frameworks and services for training and inference of large language models and transformer architectures.

• Implemented LLM distributed training, acceleration, and performance modeling using PyTorch, JAX, and Triton at a stealth AI startup.
• Designed high-throughput ML inference services with Ray and Kubernetes, including scalable Retrieval-Augmented Generation applications with LLMs and vector databases at an AI/ML company.
• Developed custom Triton kernels to accelerate LLM training and inference, alongside quantization and tensor parallelism optimizations.

3. Implement distributed training and parallelization strategies to reduce model development cycles

Explored and applied advanced parallelism techniques to enhance training efficiency across distributed systems.

• Explored model parallelism strategies to accelerate LLM training and inference using PyTorch and JAX.
• Optimized model inference through tensor parallelism, continuous batching, and quantization in production environments.
• Built MLOps orchestration frameworks with Spark data processing, multi-GPU training, and job failure recovery at a major ride-sharing company.

4. Implement tooling for data processing, model experimentation, and continuous integration

Created developer tools, pipelines, and observability for streamlined AI model development workflows.

• Developed CLI tools, workflow schedulers, and pipeline observability for model development at a major ride-sharing company.
• Built A/B testing platforms and data analytics pipelines using Spark, Kafka, Cassandra, and Redis at a global information services company.
• Researched capacity planning tools and performance modeling for LLMs, enhancing experimentation cycles.

Requirements & Candidate Alignment