> Markdown version of [/videos/397-what-non-automotive-machine-learning-projects-can-learn-from-automotive-machine-learning-projects?t=1150](https://www.wearedevelopers.com/videos/397-what-non-automotive-machine-learning-projects-can-learn-from-automotive-machine-learning-projects?t=1150). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# What non-automotive Machine Learning projects can learn from automotive Machine Learning projects

Prepare your machine learning pipelines for strict regulations like the EU AI Act. Learn how adopting rigorous autonomous driving standards keeps your models traceable, robust, and failure-proof.

- **Speakers:** Jan Zawadzki
- **Event:** World Congress 2022
- **Published:** June 15, 2022
- **Duration:** 48:01
- **URL:** https://www.wearedevelopers.com/videos/397-what-non-automotive-machine-learning-projects-can-learn-from-automotive-machine-learning-projects

## Summary

The development of machine learning for safety-critical environments like autonomous driving requires strict governance and procedural rigidity that can natively benefit non-automotive AI applications. As impending regulations like the EU AI Act prepare to classify various systems—including recommendation engines and HR algorithms—as high-risk, engineering teams must adopt robust, traceable workflows to prevent bias and operational failure. Drawing from Cariad’s strategy to build a unified software architecture for the Volkswagen Group, treating a fleet as a continuously updating AI platform requires decoupling the hardware from the operating system and applying rigorous functional safety standards like SOTIF (Safety Of The Intended Functionality).

By adapting the traditional automotive V-model into an iterative "Three V-model," teams can successfully decouple dataset creation, python-based model training, and low-level C++ embedded porting. A defining practice is treating the dataset as an independent product, governed by strict KPIs and Operational Design Domain (ODD) modeling, which explicitly defines the exact environmental limits under which a model guarantees performance. Rather than iterating endlessly without concrete constraints, developers can enforce end-to-end traceability by managing "everything as code," allowing precise root-cause analysis when an algorithm fails to meet baseline requirements.

To harden computer vision and general machine learning pipelines against real-world unpredictability, automotive engineering leverages specialized techniques for edge-case scalability. Implementing Out-of-Distribution (OOD) sampling trains models to increase entropy and recognize their own uncertainty when encountering novel objects. Additionally, utilizing composite strategies like AugMix improves model robustness and mitigates sensor data drift without exponentially inflating dataset size. Ultimately, deploying delta learning and meta-segmentation enables engineering teams to scale baseline models across vastly different domains by only adjusting for the delta of the new environment, significantly reducing retraining costs.

**Keywords:** automotive machine learning, safety-critical ai development, eu ai act compliance, operational design domain modeling, functional safety sotif, dataset as a product, out-of-distribution sampling, software development v-model, computer vision data augmentation, delta learning meta-segmentation, everything as code traceability, hardware-software decoupling, machine learning traceability, autonomous driving perception

## Chapters

1. **Introduction to safety-critical machine learning in automotive contexts** (00:05) — How statistical models safely execute high-speed decisions in vehicles.
1. **Market growth and corporate investment trends in artificial intelligence** (02:57) — Why exponential data generation drives significant enterprise investment in machine learning.
1. **Regulatory compliance and the upcoming European Union AI Act** (06:53) — How new legislation categorizes risks for both embedded systems and standard recommendations.
1. **Digitizing the vehicle of the future at Volkswagen Group** (10:31) — How consumer demand, climate crisis, and social challenges reshape automotive engineering.
1. **Decoupling hardware and software components through standardized platform architecture** (14:45) — Replacing supplier black boxes with unified operating systems to improve agility.
1. **Creating virtual feedback cycles for in-car artificial intelligence platforms** (19:10) — Leveraging vehicle fleets to continuously capture operational data for model improvement.
1. **Perception layers and trajectory planning in autonomous driving systems** (21:37) — Translating raw sensor inputs into actionable movement commands using statistical models.
1. **Function safety assurance and operational design domain modeling concepts** (23:47) — Defining specific environmental boundaries to guarantee expected vehicle performance for statistical safety models.
1. **Adapting traditional waterfall development methodologies for machine learning teams** (25:26) — Why rigid requirements gathering remains necessary before writing any data science code.
1. **Separating dataset creation from low-level software implementation steps** (28:21) — Breaking apart standard linear development to allow faster iterative training loops.
1. **Establishing comprehensive safety assurance cases for operational machine learning** (30:22) — Measuring relevant failure points and ensuring system redundancy when neural networks fail.
1. **Managing machine learning datasets as distinct internal engineering products** (32:13) — Setting concrete requirements and continuous tracking for data collection and testing.
1. **Improving computer vision resilience using augmentation and out-of-distribution sampling** (35:29) — Combining multiple synthetic image alterations to reduce sensory data drift issues.
1. **Scaling algorithm functionality between platforms with delta learning techniques** (37:54) — Adjusting existing models for new target domains without requiring complete retraining.
1. **Audience questions on data privacy, sustainability, and market hype** (40:07) — Clarifying anonymous collection methods, carbon neutral goals, and realistic development expectations.

## Related Moments

- [Introduction to machine learning in the automotive industry](https://www.wearedevelopers.com/videos/61-how-machine-learning-is-turning-the-automotive-industry-upside-down) (from "How Machine Learning is turning the Automotive Industry upside down")
- [Audience Q&A on autonomous driving models and data](https://www.wearedevelopers.com/videos/519-finding-the-unknown-unknowns-intelligent-data-collection-for-autonomous-driving-development) (from "Finding the unknown unknowns: intelligent data collection for autonomous driving development")
- [Navigating automotive complexity with AI runtime environments](https://www.wearedevelopers.com/videos/198-developing-an-ai-sdk) (from "Developing an AI.SDK")
- [Addressing participant questions on liability and machine learning](https://www.wearedevelopers.com/videos/374-the-future-of-automotive-mobility-upcoming-e-e-architectures-v2x-and-its-challenges) (from "The future of automotive mobility: Upcoming E/E architectures, V2X and its challenges")
- [The virtuous cycle of machine learning in connected cars](https://www.wearedevelopers.com/videos/61-how-machine-learning-is-turning-the-automotive-industry-upside-down) (from "How Machine Learning is turning the Automotive Industry upside down")
- [Industrial applications of machine learning in autonomous vehicles](https://www.wearedevelopers.com/videos/260-getting-started-with-machine-learning) (from "Getting Started with Machine Learning")

## Related Articles

- [How software is steering vehicle technology](https://www.wearedevelopers.com/magazine/515-how-software-is-steering-vehicle-technology)
- [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development)
- [MLOps – What’s the deal behind it?](https://www.wearedevelopers.com/magazine/125-mlops-what-s-the-deal-behind-it)
- [Stephan Gillich - Bringing AI Everywhere](https://www.wearedevelopers.com/magazine/489-stephan-gillich-bringing-ai-everywhere)

## Related Jobs

- [Data Scientist](https://www.wearedevelopers.com/jobs/ext/1351648-data-scientist) at **Almedia**
- [Principal Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/1706410-principal-machine-learning-engineer) at **Almedia**
- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Staff Software Engineer, Copilot Experiences](https://www.wearedevelopers.com/jobs/ext/164361-staff-software-engineer-copilot-experiences) at **GitHub**
- [Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/1597388-machine-learning-engineer) at **ZEISS Group**
- [Security Architect - AI](https://www.wearedevelopers.com/jobs/ext/1581899-security-architect-ai) at **ZEISS Group**