Modernizing DoorDash Model Serving Platform with Ray Serve

At DoorDash, ML is applied to solve many customer facing problems. For serving model predictions, the ML Platform team initially built a highly performant prediction service that serves several models in production. Initially the model frameworks supported were limited to PyTorch and LightGBM models. With the fast pace of innovation in the field of ML, the focus has shifted to more flexibility and developer velocity. We have had more interest in NLP and image processing use cases where there is a large variety of libraries that we aimed to support.
The ML platform team has recently adopted Ray for model training and inference. In this talk, we will share how we built a new model serving platform keeping flexibility and self-service as core ethos. We have evaluated many different frameworks and found that the data scientist friendly approach that Ray Serve offers is a great match for what we were looking for. We will tell the story of how we shifted our prediction services from our earlier generation to the Ray Serve ecosystem.
Find the slide deck here: drive.google.c...
About Anyscale
---
Anyscale is the AI Application Platform for developing, running, and scaling AI.
www.anyscale.com/
If you're interested in a managed Ray service, check out:
www.anyscale.c...
About Ray
---
Ray is the most popular open source framework for scaling and productionizing AI workloads. From Generative AI and LLMs to computer vision, Ray powers the world’s most ambitious AI workloads.
docs.ray.io/en...
#llm #machinelearning #ray #deeplearning #distributedsystems #python #genai

[@binaryutopia7946]

Great presentation. Great to see how ML models are heavily used almost in each stage of an order