GenAI/Applied AI

What matters most

• Build usable engineering AI workflows

  The role builds and maintains RAG pipelines, prompt and agent workflows, and evaluation harnesses for internal engineering use cases. Prepare to explain how you would develop and iterate on a Python-based workflow using an LLM API.

• Make outputs measurable and reliable

  The posting emphasizes checking pipeline runs, triaging quality regressions, reducing hallucinations through knowledge-graph grounding, and extending evaluation cases. Prepare a concrete approach to measuring accuracy and hallucination rates and using results to guide changes.

• Prototype ASPICE-supporting artifacts

  The advertised work includes AI-assisted requirements drafting, traceability, and test-case generation, including an ASPICE SOW-to-artifact use case. Prepare to discuss how you would test generated artifacts against real project data with a systems software engineer and incorporate feedback; ASPICE awareness is a plus, not a required compliance qualification.

• Turn experiments into stakeholder decisions

  The role supports fine-tuning and domain-adaptation experiments, documents findings, maintains a prototype backlog, and demonstrates work to engineering and product stakeholders. Prepare to show how you would record results and communicate what a prototype does and does not establish.

Likely challenges to think through

• Balancing artifact coverage with trustworthy output

  Preparation hypothesis: generating more requirements or test cases may increase apparent coverage while introducing unsupported or incorrect content. Prepare to explain how grounding, project-data checks, and targeted evaluation cases would expose that trade-off.

• Diagnosing regressions across a changing pipeline

  Preparation hypothesis: when RAG retrieval, prompts, agent steps, and model adaptation all affect outputs, a quality drop may be hard to attribute. Prepare a method for logging runs, comparing evaluation cases, and changing one component at a time.

• Prioritizing exploratory AI+X work

  Preparation hypothesis: requests spanning engineering, business development intelligence, and HR automation could compete for prototype time. Prepare to describe how you would use stakeholder feedback, documented findings, and the prototype backlog to compare use cases without treating early demos as validated solutions.

Topics to prepare

• NLP and deep learning fundamentals

  Required foundation. Revise how language models represent text, generate outputs and learn from training data; explain how these concepts affect an automotive requirements task.

• RAG and grounded generation

  RAG familiarity is required, and the role builds RAG pipelines. Sketch how retrieved project material would support a drafted requirement, then identify where unsupported claims could enter.

• Prompt and agent-workflow concepts

  The posting requires prompt-engineering and basic agent-framework understanding. Compare a single prompted draft with a multi-step draft-and-check workflow, including how each step would receive context.

• Evaluation and hallucination reduction

  The role measures accuracy and hallucination rates and explores knowledge-graph grounding. Define examples of correct, incorrect and unsupported outputs, then explain how grounding could help check them.

• Fine-tuning and domain adaptation

  The role supports experiments for automotive requirements and compliance documentation. Outline when adapting a model might be useful versus improving retrieval, and what evaluation would show whether it helped.

• Requirements artifacts and ASPICE terminology

  The tools support requirements drafting, traceability and test-case generation in an ASPICE pipeline. Map a sample statement of work to a requirement and a test case. ASPICE terminology is a plus; the posting does not require domain-standard compliance experience.

Skills to revise

• Python and LLM API prototyping

  Strong Python skills and familiarity with LLM APIs are required. As practice, write a small Python script that submits a requirements-drafting prompt through an API, records the response and handles a failed call.

• Building a RAG pipeline

  The role builds and maintains RAG pipelines. As practice, create a small example using engineering-style documents: retrieve relevant passages, pass them to a model and retain the passages alongside its answer.

• Iterating prompts and agent workflows

  The role builds prompt and agent workflows. As practice, prototype a draft-then-check sequence for a requirement, and compare its output with a single-prompt version.

• Preparing data and testing adaptation

  Dataset preparation and fine-tuning or domain-adaptation experiments appear in the typical day. As practice, assemble a small set of example requirements, separate training examples from evaluation cases and state what change an adaptation experiment should measure.

• Extending an evaluation harness

  The posting describes adding cases, logging results and triaging quality regressions. As practice, build a small test set with expected outputs, score accuracy and unsupported claims, then investigate one failed case.

• Testing artifacts with engineering feedback

  The role tests generated requirements and test cases against real project data with a systems software engineer. As practice, review a sample generated requirement, trace it to its source, draft a test case and record specific revisions a reviewer might request.

• Gather evidence for the core skills

  Prepare brief examples of Python work, NLP or deep-learning fundamentals, and any LLM API, RAG, prompt or agent-workflow experience. Note which examples are hands-on and which are conceptual.

• Prepare a RAG project walkthrough

  Bring a diagram or concise walkthrough covering document ingestion, retrieval, prompting, an LLM API call and failure handling. Use an existing project if available; otherwise label it as a proposed design for internal engineering documents.

• Design a requirements-to-test-case prototype

  Sketch how an AI tool could turn a statement of work into draft requirements, traceability links and test cases. Show where source evidence and human review enter the flow; learn enough ASPICE terminology to explain the pipeline without claiming compliance expertise.

• Draft an evaluation and regression plan

  Prepare sample evaluation cases and a results table for accuracy, unsupported claims and hallucination rate. Explain how you would inspect a failed pipeline run, add a regression case and compare the next run.

• Compare reliability and adaptation options

  Outline when knowledge-graph grounding might help, what data a domain-adaptation or fine-tuning experiment would need, and how you would evaluate whether either approach improves results.

• Plan a feedback-driven demo

  Prepare a short prototype demo outline and a way to record systems-engineer feedback on generated requirements or test cases. Include how findings would become backlog items for engineering and product stakeholders.

• Write questions for the employer

  Ask which ASPICE statement-of-work-to-artifact use case is the initial priority, what project data and review support are available, how quality is measured, and how the lead prioritises engineering work against other AI+X ideas.