AI/ML Engineer – LLM & Voice Pipeline (w/m/d) - hands on experience in hands-on AI/ML, LangChain/LangGraph, Azure, ASR engine

Luxoft Germany; Ingolstadt; Vor Ort; EUR 70.000 - 95.000; Vollzeit

Zusammenfassung

Luxoft Germany is advancing its in-vehicle voice assistant with AI-powered capabilities integrated into the MIB3 infotainment system and Azure OpenAI services. The team develops end-to-end data pipelines, on-device AI, and edge inference within a joint product team collaborating with client engineers.

We seek an experienced AI/ML engineer with strong Python skills, expertise in LLM frameworks, and hands-on RAG systems to help build robust, safe dialogue and audio processing modules for

Qualifikationen

• 3+ years of hands-on AI/ML engineering in production environments.
• Strong Python skills with experience in LLM frameworks (LangChain, LangGraph).
• Experience building RAG systems (embeddings, chunking, retrieval eval).
• Familiarity with Azure OpenAI Service or OpenAI API; knowledge of prompt engineering.

Aufgaben

• Design, implement, and optimize RAG pipelines with Azure AI Search and LLM endpoints.
• Develop ASR and TTS integration modules with pyAudio and librosa.
• Create LLM prompts, evaluation frameworks, and safety guardrails for in-vehicle dialogue.
• Prototype few-shot adapters and intent classification; transition to production.
• Collaborate with Backend Engineers to connect AI layer with streaming infra (FastAPI/Kafka).
• Contribute to edge AI components including on-device inference and latency management.
• Monitor model quality metrics like BLEU, WER, CER and faithfulness.

Kenntnisse

Python

LLM frameworks (LangChain/LangGraph)

RAG systems

Azure OpenAI / OpenAI API

ASR & TTS integration

English (B2+)

Tools

PyAudio

librosa

FastAPI

Kafka

Azure Kubernetes Service (AKS)

Jobbeschreibung

Our client is advancing its in-vehicle voice assistant into an intelligent, AI-powered companion. Since 2024, they have been incorporating large-language-model capabilities (Azure OpenAI / ChatGPT) into vehicles equipped with the MIB3 infotainment system, with new E³-architecture models featuring enhanced voice functions from the factory. The AIME (AI Model Engine) backend program supports this development over a multi-year timeline, addressing natural-language dialogue, empathic communication, and the complete cloud-edge data pipeline. DXC Luxoft serves as the end-to-end delivery partner, collaborating closely with the client’s engineers within a joint product team on the Azure platform (AKS, Azure OpenAI, Managed Identity, Azure Monitor, Azure DevOps).

Responsibilities

• Design, implement, and optimize RAG pipelines integrating Azure AI Search, vector stores, and LLM completion endpoints
• Develop and maintain ASR and TTS integration modules, including audio pre- and post-processing using pyAudio and librosa
• Create LLM prompt chains, evaluation frameworks, and safety/alignment guardrails for in-vehicle dialogue scenarios
• Prototype AI capabilities such as few-shot adapters, intent classification, and empathic dialogue, then transition prototypes to production
• Collaborate with Backend Engineers to develop the FastAPI/Kafka interface connecting the AI layer with the streaming infrastructure
• Contribute to edge AI components including on-device inference, model quantization, and latency management for in-vehicle use
• Monitor and enhance model quality metrics including BLEU, WER, CER, and faithfulness via continuous evaluation pipelines.
• Participate in code reviews with cross-functional teams, including customer ML engineers.

Skills

Must have

• 3+ years of hands-on AI/ML engineering in production environments
• Strong Python skills; experience with LLM frameworks: LangChain, LangGraph, or equivalent
• Practical experience building RAG systems (chunking strategies, embedding models, retrieval evaluation)
• Familiarity with Azure OpenAI Service or OpenAI API; prompt engineering best practices
• Experience with at least one ASR engine (Whisper, Azure Speech, or equivalent) and a TTS system
• English B2 or above

Nice to have Skills

• Automotive in-vehicle speech processing context (noise cancellation, far-field microphones)
• Experience with LLM evaluation frameworks (DeepEval, Ragas, PromptFlow).