Master Thesis in Extending GEMM for Time-series DSP Algorithms

Robert Bosch Group; Renningen; Vor Ort; EUR 40.000 - 60.000; Vollzeit

Zusammenfassung

A leading company in technology innovation seeks candidates for a master thesis project focused on extending GEMM for DSP algorithms related to time-series data analysis in acoustic signals. The candidate will engage in cutting-edge research, optimizing algorithms for hardware implementation while leveraging interests in future technologies.

Qualifikationen

• Master's studies in a relevant field required.
• Experience in Digital Design, Verilog/VHDL, and Python essential.
• Fluent in English, German is a plus.

Aufgaben

• Develop and optimize algorithms for GEMM-based accelerators.
• Investigate input data representations for acoustic scene analysis.
• Design hardware considerations for processing chains and neural networks.

Kenntnisse

Digital Design

System Verilog/VHDL

Python

Ausbildung

Master's studies in Electrical Engineering

Master's studies in Computer Science

Jobbeschreibung

Master Thesis in Extending GEMM for Time-series DSP Algorithms

Position: Full-time

Company: Robert Bosch GmbH

At Bosch, we shape the future by inventing high-quality technologies and services that spark enthusiasm and enrich people’s lives. Our promise to our associates is to grow together, enjoy our work, and inspire each other. Join us and feel the difference.

We are looking forward to your application!

Thesis Description

Prior to feeding data to neural networks, spectrum is typically generated using sliding windows FFT and MFCC on acoustic signals. This approach treats acoustic signals as images, and image-based neural networks such as CNNs are utilized for tasks like keyword spotting and denoising.

Extracting temporal and frequency information from the spectrum requires heavy pre-processing. You will develop and optimize algorithms to utilize the gains provided by GEMM-based accelerators, exploring different approaches to exploit features present in acoustic signals. Leveraging time encoding neural networks, you will aim to better present the time-series characteristics of acoustic signals with lightweight pre-processing.

You will also investigate various input data representations and DSP-heavy pre- and post-processing techniques to analyze acoustic scenes, enabling efficient mapping on GEMM-based accelerators. Hardware design considerations will be central in designing and optimizing processing chains, including neural network design to facilitate hardware implementation.

Qualifications

• Education: Master's studies in Electrical Engineering, Computer Science, or a comparable field
• Experience and Knowledge: Digital Design, (System)Verilog/VHDL, Python; background in Neural Networks
• Personality and Working Practice: Organized, able to work independently
• Enthusiasm: Keen interest in future technologies and trends, with a passion for innovation
• Languages: Fluent in English; German is a plus

Start: According to prior agreement

Duration: 6 months

Requirement: Enrollment at university. Please attach your CV, transcript of records, examination regulations, and if applicable, a valid work and residence permit.

Diversity and inclusion are integral to our culture. We welcome applications from all genders, ages, disabilities, religions, ethnic origins, and sexual identities.

Further Information

For more information about the job, contact:

Andre Guntoro (Functional Department)
[phone]