Open to opportunities

Jacob Lewis

@jacoblewis

Message

I build efficient, production-ready generative AI systems for multimodal applications.

United States

Message

What I'm looking for

I seek senior roles building production-grade multimodal generative AI—optimizing models for edge and cloud, leading cross-functional teams, prioritizing performance, scalability, safety, and open-source collaboration in mission-driven organizations.

I am a senior generative AI engineer who designs, builds, and deploys multimodal models across image, audio, and video, with recent contributions to Stable Diffusion 3, Stable Audio 2.0, and Stable Video 4D. I focus on boosting model quality, reducing inference latency, and making models practical for both edge and cloud deployment.

I optimize models for ARM/mobile and GPU/cloud, lead integrations into production pipelines, and champion open-source releases that drive community adoption and enterprise impact. My work balances research and engineering—delivering scalable, safe, and resource-efficient systems informed by prior roles at StabilityAI, Cohere, and Hugging Face, and grounded by advanced AI training from MIT.

Experience

Work history, roles, and key accomplishments

Current

Generative AI Engineer

Current

Stability AI

Jan 2024 - Present (1 year 7 months)

Led rollout of Stable Audio 2.0 achieving 30% faster inference and 25% higher adoption, engineered a 60%-smaller on-device ARM model that expanded smartphone users by 15M, and piloted Stable Video 4D and Stable Fast 3D to cut prototype generation time by 70% and increase API usage by 50%. Co-engineered Stable Diffusion 3 improvements that boosted multi-subject prompt accuracy 30% and reduced consu

Latent Diffusion Audio Generation Model Compression ARM CPU Optimization TensorRT Onnx PyTorch

Senior Machine Learning Engineer

Cohere

Jan 2022 - Dec 2023 (1 year 11 months)

Engineered fine-tuning and deployment pipelines for Command R, raising summarization accuracy to 80.2% while cutting inference costs by 85% and reducing compute spend 3×; built RAG-powered Compass and North to lower hallucination rates by 45%. Launched Aya multilingual models and hybrid model merging pipelines to accelerate releases 40% and enable secure private-cloud deployments across finance an

Hybrid Model Merging Azure CI CD Vector Search

Machine Learning Engineer

Hugging Face

Jun 2020 - Dec 2021 (1 year 6 months)

Scaled the Datasets library (+225% submissions) and optimized the Tokenizers Rust pipeline to accelerate processing 4× and reduce memory footprint 60%, enabling large-scale long-context modeling and enterprise inference adoption. Built MoE support and end-to-end dataset-to-model pipelines that increased inference efficiency and enabled widespread user deployments.

Rust Tokenization Mixture Of Experts Long Context Modeling Datasets PyTorch Transformers

Software Engineer

InVisionApp, Inc

Jul 2019 - Jun 2020 (11 months)

Optimized canvas and animation pipelines with GPU-accelerated rendering to reduce rendering lag 60% and enable designers to prototype up to 4× faster; expanded mobile mirroring features boosting mobile prototype testing 45%. Resolved 120+ critical bugs and improved platform uptime to 99.8%, raising customer satisfaction by 15%.

GPU Accelerated Rendering JavaScript Performance Optimization Mobile Integration Frontend Engineering Debugging

AI Research Intern

OpenAI

Jun 2018 - Aug 2018 (2 months)

Implemented and benchmarked the initial 117M-parameter transformer (GPT-1), improving pre-training throughput 2.5× via optimized PyTorch pipelines and enabling downstream fine-tuning gains across multiple NLP tasks. Contributed to RL training and sim-to-real tooling that increased simulation efficiency and improved real-world robotic manipulation success rates.

PyTorch Domain Randomization