Open Audio ML Research (Master Thesis / Internship)

Embodied AI; Zürich; Hybrid; CHF 20.000 - 27.000; Teilzeit

Benefits dieser Stelle

Paid Master thesis

Own Topic

Real Attack Data

Weekly Supervision

Hybrid setup

Zusammenfassung

Aurigin.ai, a Zurich-born startup, invites MSc students to propose an audio ML thesis or internship. We seek a concrete idea in audio, speech, or signal processing, with strong Python and ML skills, ideally PyTorch, and fluency in English.

You'll work with real attack data and a detector running in production, constrained to under 50 ms latency and a measured false alarm rate. Expect weekly supervision and a flexible hybrid setup with two days in Zurich.

Qualifikationen

• Pursuing or recently finished MSc in CS/EE/ML.
• Concrete idea in audio, speech, or signal processing.
• Strong Python and ML, ideally PyTorch.
• Willing to argue for your idea and iterate.

Aufgaben

• Send a proposal detailing the question, significance, approach, and success criteria.
• Shape the scope and feasibility with the team, settle on a final version.
• Run experiments, ablations, and honest evaluation with supervision.
• Test against real conditions using production-like data and latency limits.
• Deliver a thesis/report and code compatible with our stack.

Kenntnisse

Python

PyTorch

Audio processing

English

Ausbildung

MSc in CS/EE/ML

Tools

PyTorch

Jobbeschreibung

Bring us the audio ML problem you want to spend six months on

About us

Aurigin.ai is a Zurich-born startup on a mission to restore trust in digital communication by protecting high-stakes organizations from AI-generated voice fraud in real time. Our deepfake detection is used to stop account takeovers, executive impersonation, and fabricated recordings before they cause damage.

We are a team of engineers, researchers, and builders with backgrounds across tech, consulting, and startups. We thrive on collaboration, embrace a "move fast to launch and iterate" mentality, and share a vision of a safer, more transparent world in the age of AI.

Our other thesis positions come with a defined problem. This one does not. It exists for people who already know what they want to work on, and it sits alongside the specific topics rather than replacing them.

Why we need you

The topics we write ourselves come from problems we already know we have. The more interesting ones tend to come from somewhere else: a paper we have not read, a technique borrowed from another field, a question that only looks obvious to someone who has been living inside it. We would rather hear one of those than have you pick from our list. What we can offer in return is the part that is hard to get from inside a university, namely recordings of real attacks, a detection engine running on live traffic, and constraints sharp enough to turn a loose research question into a precise one: a verdict has to land in under 50 ms, and a false alarm costs about as much as a miss. If your idea fits in there, the topic is yours.

What you will do

• Send us a proposal: the question, why it matters, roughly how you would attack it, and what would count as an answer. One or two pages is plenty
• Shape it with us: we will push hard on scope, feasibility, and how it connects to what we run in production, then settle on a version we both want to see finished
• Run the research: experiments, ablations, and honest evaluation, with weekly supervision and a lot of freedom in between
• Test it against real conditions: our data, our latency and false-alarm budget, and audio that has been through real devices, rooms, and phone lines
• Land it: a thesis or report, and where the result holds up, code that goes into our stack

About you

• Pursuing or recently finished an MSc (or equivalent) in CS, EE, ML, or a related field
• A concrete idea you want to work on, and enough grounding in audio, speech, or signal processing to explain why it is worth doing
• Strong Python and hands-on ML, ideally PyTorch
• Willing to argue for your idea and then let it change: we will challenge the proposal, and the best version of it is rarely the first one
• Rigorous about measurement, because a result you cannot evaluate is not a result
• Fluent in English, available for around 6 months, and able to be in the Zurich office at least two days a week

Nice to have

• Published, preprinted, or open-sourced work in audio, speech, or ML
• A university supervisor already lined up, if this is a Master thesis
• Experience with generative audio, deepfake detection, speaker verification, or real-time inference
• A track record of finishing self-directed projects without someone setting the milestones

Benefits

• Paid Master thesis or internship
• Your own topic, chosen by you and taken seriously rather than handed down
• Access to real attack data and a detector running in production, which is hard to come by as a student
• Weekly supervision from people who work on this full-time, without being told what to do day to day
• Flexible hybrid setup: work where you are most productive, with at least two days a week in the Zurich office

Explore Acoustic Atlas