PhD Position (f/m/d) - Hardware/Software Co-Design for Efficient On-Device Incremental Learning on Embedded NPU Platforms

NXP Semiconductors; Hamburg; Vor Ort; EUR 18.000 - 29.000; Vollzeit

Zusammenfassung

NXP Semiconductors in Germany is seeking a PhD candidate to explore Hardware/Software Co-Design for Efficient On-Device Incremental Learning on Embedded NPU Platforms. The project targets incremental learning on MCU/MPU-class platforms with integrated NPUs under strict memory, latency, energy, and compute limits.

You will develop new methods, runtimes, and system optimizations, and evaluate approaches on real embedded hardware.

Qualifikationen

• Master’s degree in Electrical Engineering, Computer Engineering, Computer Science, Embedded Systems, AI, or related field.
• Strong background in machine learning / deep learning, embedded systems, and computer architecture.
• Programming experience in Python and C/C++.
• Hands‑on experience with PyTorch or TensorFlow.
• Very good written and spoken English.

Aufgaben

• Conduct research on incremental learning for embedded AI systems.
• Develop and implement algorithms, runtime mechanisms, and system optimizations for on-device adaptation.
• Evaluate solutions on real embedded hardware platforms with NPUs.
• Analyze trade-offs across accuracy, forgetting, energy, latency, and memory footprint.
• Publish results in leading conferences and journals.
• Collaborate closely with experts in machine learning, embedded software, hardware architecture, and AI acceleration.

Kenntnisse

machine learning / deep learning

embedded systems

computer architecture

AI accelerators / NPUs

compilers or runtime systems

Python

C/C++

PyTorch

TensorFlow

English language proficiency

Ausbildung

Master’s degree in Electrical Engineering, Computer Engineering, Computer Science, Embedded Systems, AI, or related field

Tools

PyTorch

TensorFlow

Jobbeschreibung

Embedded AI is evolving beyond static inference toward efficient, privacy-preserving, and adaptive on-device intelligence. As embedded platforms increasingly integrate dedicated AI accelerators such as NPUs, new opportunities arise to enable learning and personalization directly on resource-constrained devices.

We are looking for a PhD candidate to work on Hardware/Software Co-Design for Efficient On-Device Incremental Learning on Embedded NPU Platforms and help drive the next wave of intelligent embedded systems.

This PhD project focuses on enabling incremental learning directly on MCU/MPU-class platforms with integrated NPUs , under strict constraints in memory, latency, energy, and compute efficiency.

The goal is to develop new methods, runtimes, and system-level optimizations that allow embedded AI systems to adapt over time in a practical and scalable way.

In this role, you will focus on the following research directions:

• Incremental learning under concept drift
  Develop methods that allow embedded AI models to adapt to changing data distributions while avoiding catastrophic forgetting.
• Parameter-efficient incremental adaptation
  Investigate lightweight update mechanisms that avoid full retraining and reduce compute and memory overhead.
• NPU-aware incremental learning runtime
  Design runtime, scheduling, and execution strategies for efficient incremental learning on heterogeneous embedded platforms with CPU/NPU resources.
• Privacy-preserving incremental personalization
  Explore local adaptation and personalization techniques that keep sensitive data on device.

You will work on the intersection of machine learning, embedded systems, runtime design, and AI hardware architecture , with the aim of enabling practical and efficient on-device learning for future edge-AI products.

In this PhD, you will:

• Conduct research on incremental learning for embedded AI systems
• Develop and implement algorithms, runtime mechanisms, and system optimizations for on-device adaptation
• Evaluate solutions on real embedded hardware platforms with NPUs
• Analyze trade-offs across accuracy, forgetting, energy, latency, and memory footprint
• Publish results in leading conferences and journals
• Collaborate closely with experts in machine learning, embedded software, hardware architecture, and AI acceleration

Your Profile:

• A Master’s degree in Electrical Engineering, Computer Engineering, Computer Science, Embedded Systems, Artificial Intelligence, or a related field
• A strong background in one or more of the following areas:
  • machine learning / deep learning
  • embedded systems
  • computer architecture
  • AI accelerators / NPUs
  • compilers or runtime systems
• Programming experience in Python and C/C++
• Hands‑on experience with machine learning frameworks such as PyTorch or TensorFlow
• Strong analytical and problem-solving skills, with the ability to work independently on challenging research topics
• Very good written and spoken English

The following experiences would be considered a plus:

• incremental / continual / online learning
• model compression, quantization, or pruning
• embedded AI deployment on MCU/MPU platforms
• NPU programming and performance analysis
• privacy-preserving machine learning
• benchmarking and profiling of AI workloads on real systems

What you can expect:

At NXP, you will work on a highly relevant PhD topic at the intersection of AI, embedded systems, and hardware/software co-design. You will have access to modern embedded AI platforms, development tools, and a collaborative environment connecting research with real product challenges.

You will work closely with experts across hardware, software, architecture, and machine learning, while contributing to technologies that can shape the future of adaptive edge intelligence.

This is an excellent opportunity for a motivated researcher who wants to combine scientific depth with strong industrial relevance.

Please note: The successful candidate may/will be responsible for security related tasks. The assignment may/will be in scope of security certifications, therefore a conscious and reliable way of working is necessary.

More information about NXP in Germany...