MLOps at Volvo Cars // Leonard Aukea // MLOps Meetup #84

MLOps Community Meetup #84! Last Wednesday we talked to Leonard Aukea, Head of Machine Learning Engineering & Operations of Volvo Cars.

//Abstract
After spending most of his career as a full-stack Data Scientist/ML Engineer, Leonard has shifted focus towards MLOps. Leonard is currently driving machine learning engineering and operations at Volvo Cars. In particular, how to effectively reap its benefits at an enterprise scale.

Leonard introduces Volvo Cars ML stack and related work focused on stitching these services together in order to reduce friction in the ML value stream. Furthermore, we discuss some learnings along the way and general thoughts around what it takes to lay a solid ML foundation in a company like Volvo Cars.

// Bio
Leonard is driving ML Engineering and Operations at Volvo Cars. He is responsible for defining the overall mission and strategy for ML Engineering and Operations, leading the build of reproducible ML systems. Leonard Aukea has spent most of his career as a Data Scientist/ML Engineer.

Leonard lives in Gothenburg, Sweden with his wife and two kids.

// Related links
Rules of Machine Learning: Best Practices for ML Engineering by Martin Zinkevich - https://developers.google.com/machine-learning/guides/rules-of-ml

Engineering best practices for Machine Learning - https://se-ml.github.io/practices/

Effective testing for machine learning systems by Jeremy Jordan - https://www.jeremyjordan.me/testing-ml/

----------- ✌️Connect With Us ✌️-------------
Join our Slack community: https://go.mlops.community/slack
Follow us on Twitter: @mlopscommunity
Sign up for the next meetup: https://go.mlops.community/register
Catch all episodes, Feature Store, Machine Learning Monitoring and Blogs: https://mlops.community/

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Leonard on LinkedIn: https://www.linkedin.com/in/leonardaukea/

Timestamps:
[00:00] Introduction to Leonard Aukea
[03:00] Driving ML Engineering and Operations
[03:05] Purpose
[03:51] Approach
[05:16] Dedicated ML Clusters
[07:20] Tecton's CI/CD pipeline + Tekton Hub + @GitHub
[07:42] Workflow
[09:42] The Template
[10:39] profiles.yaml
[11:52] Onboarding CI -Tekton dasgboard
[12:25] /workflows
[13:05] Pipeline versioning
[13:28] Commit CI - KubeFlow pipelines
[14:21] Trigger pipeline with github comment
[15:14] /run pipeline example_pipeline.py
[15:17] Versioning Models/Data
[16:56] Versioning
[18:32] Continuous Monitoring
[20:32] Is it enough for ML?
[20:47] Roadmap
[23:59] Challenges
[24:05] Having the right skills?
"You can't be an AI expert these days and not have some grounding in software engineering." - Grady Booch
[24:20] Data scientists + @GitHub
[25:29] Coding skills
[25:54] Data Scientist + Git + CI/CD + Container Basics + SW/ML Testing
[27:00] "Do machine learning like the great engineer you are, like not the great machine learning expert you aren't." - Martin Zinkevich
[27:25] Treat ML-like our cars
[27:40] SW vs ML
[27:49] Software 1.0
[28:04] Conventions
[28:16] Software 2.0
[28:33] Evaluation
[29:00] Compare performance between models and make relative judgments
[29:05] Conventions
[29:15] Testing vs Evaluation
[29:52] Terminology
[30:21] Revised Conventions
[32:13] The Red Team
[34:23] MLOps is an engineering practice
[34:52] The Bibles
[35:32] "Rule #4: Define the operational domain for your ML system." - Svet Penkov
[36:12] Timeline for Volvo MLOps system development and components
[37:33] Volvo is hiring!
[38:31] Hard challenges
[40:49] Staing up to date, security flaws and dependencies
[42:31] Kubeflow on Google Cloud deployment
[43:48] Self-driving cars
[45:46] Volvo's organization
[47:16] Volvo's data flowing
[48:28] Design decisions
[49:55] Centralized vs Decentralized
[52:23] Convincing data scientists
[55:24] Blog post suggestion inside the company on open-source mindset

Видео MLOps at Volvo Cars // Leonard Aukea // MLOps Meetup #84 канала MLOps.community