AI/ML Intern, Long-Term Memory & Multimodal SLMs for Automotive Agents

At Mercedes-Benz Research & Development North America (MBRDNA), we are committed to delivering world-class automotive technologies that push the boundaries of what is possible. Our teams of highly skilled engineers and designers use cutting-edge software and technology to enhance the driving experience and reduce environmental impact.

Position Overview

This internship will start in January 2027. The Mercedes-Benz AI Experiences department is seeking an AI/ML Intern for Spring and Summer 2027 to join the MB.Intelligence Context team. You will build on-device long-term memory for automotive agents using small language models (SLMs), Liquid AI's multimodal LFM3-Omni model, and trigger-based activation to help deliver personalized, privacy-preserving experiences in Mercedes-Benz vehicles.

Responsibilities

• Design and implement multimodal long-term memory (RAG) for automotive agents in resource-constrained in-vehicle environments.
• Fine-tune LFM3 models, especially LFM3-Omni-2B, and other small language models for memory, preference extraction, and function calling using SFT, LoRA, and QLoRA.
• Apply reinforcement learning such as DPO and GRPO to improve memory, activation, and tool-use decisions.
• Build trigger-based activation from multimodal cabin and vehicle signals using decision models and lightweight classifiers.
• Build classifiers for turn detection, PII detection, and intent routing.
• Optimize and profile models for on-device deployment through quantization and distillation.
• Build evaluation frameworks for memory, retrieval, and trigger accuracy, and present findings to Mercedes-Benz teams.

Minimum Qualifications

• Master's or higher degree in computer science, computer engineering, electrical engineering, mathematics, statistics, or a related field.
• Python and PyTorch experience with Git-based workflows.
• Experience fine-tuning small language models using Hugging Face Transformers, TRL, PEFT, LoRA, or QLoRA.
• Familiarity with reinforcement learning for language models, multimodal models, decision models, classifiers, RAG, embedding models, vector search, model optimization, and ML evaluation methodology.
• Strong collaboration and written communication skills.

Preferred Qualifications

• Hands-on experience with Liquid AI models such as LFM2.5 and LFM3.
• Edge LLM inference using llama.cpp, LiteRT, ExecuTorch, or ONNX Runtime.
• Privacy-preserving machine learning, agentic AI/MCP, or Android embedded systems experience.
• Always-on perception experience, including wake-word, voice activity, or audio/visual event detection.

This position may involve access to export-controlled technology, hardware, software, and technical data subject to United States export regulations.

Benefits

Full-time employee benefits include medical, dental, and vision insurance, a 401(k) with employer match, company-paid holidays, paid time off, tuition assistance, and wellness or fitness reimbursement programs. Internships and contractors are excluded from full-time employee benefits.

MBRDNA is an equal opportunity employer and strongly supports diversity in the workforce.