Applied AI Scientist Intern - Autonomous Confidence Calibration
Company
At Subduxion, we're a dynamic startup on a mission to create real value for organizations through cutting-edge technological advancements, with a strong focus on Agentic AI. We're building intelligent systems that go beyond traditional AI, empowering businesses to automate complex processes and make smarter decisions.
Located at the heart of innovation on the High Tech Campus - Europe's smartest square kilometer - we're a small, entrepreneurial team that's all about collaboration, risk-taking, and authentic relationships. As a startup, we're flexible, fast-paced, and full of opportunities for you to make an immediate impact. If you're passionate about turning data into actionable AI solutions and bridging the gap between client needs and tech implementations, this is your chance to grow with us in the age of AI.
Who are we looking for?
We're opening 3 internship positions for Applied Scientists who want to do real AI/ML research with direct product impact. You'll work on the intelligence layer of our AI SDR (Sales Development Representative) platform - specifically on the core question behind its evolution: when is the AI confident enough to act on its own, and when should it defer to the human?
Our platform is meant to become the EU's most capable AI SDR, helping founder-led B2B businesses sell better using AI. Get the confidence boundary right and the AI becomes a trusted autonomous agent. Get it wrong and users lose trust. Your research defines that boundary.
We're building toward full sales autonomy by end of 2026. Your work directly shapes whether we get there. This isn't a paper-writing exercise: your models ship to production.
These are founding research roles. Perform well, and there's a direct path to Head of Applied Research at Subduxion, leading the AI research function across all our products.
Responsibilities
Collaborate with our team to design, build, and deploy Agentic AI models that solve real-world business problems, from data preprocessing to model training and integration.
- Research and develop the framework that determines when the AI should act independently versus when it should ask for human confirmation.
- Design calibration methods that measure and improve the reliability of the AI's own confidence estimates across different tasks and contexts.
- Build prototypes, run experiments, and ship what works into production alongside our engineering team.
- Publish internal research reports and present findings to stakeholders.
- Stay current on relevant academic literature and translate promising techniques into practical implementations.
- Ensure all approaches meet GDPR/AVG requirements by design.
We keep things flexible: remote-friendly hours mean you can balance work with life, and as a startup, we're not bogged down by bureaucracy. You'll have the freedom to experiment and learn quickly.
Minimum Qualifications
We're keeping our bar realistic for (entry-level) talent:
- Pursuing or recently completed a Master's degree in AI, Machine Learning, Computer Science, Data Science, Applied Mathematics, Econometrics, or similar.
- Solid machine learning fundamentals - you can reason about models, not just run notebooks.
- Programming proficiency in Python with hands-on ML framework experience (PyTorch, TensorFlow, scikit-learn).
- Ability to read and implement research papers.
- Strong communication skills in English. Dutch is a plus.
- Entrepreneurial minded, which means that you ship, you iterate, you take ownership and ring the bell on-time.
No need for years of experience; if you've just graduated or have 0-2 years under your belt from internships or junior roles, that's perfect. We're more interested in your potential and entrepreneurial mindset.
Preferred Qualifications
These would make you stand out, but they're not deal-breakers - we're a startup, so we're betting on growth:
- Experience with uncertainty quantification, calibration, or Bayesian methods.
- Familiarity with reinforcement learning from human feedback (RLHF) or human-in-the-loop systems.
- Exposure to large language models in applied settings - especially agentic patterns.
- Understanding of AI safety, alignment, or trustworthy AI concepts.
- A track record of building things - e.g. hackathons, open-source, side projects with real implementation.
- Understanding of GDPR/AVG and ethical AI.
- That entrepreneurial spark: maybe you've started a personal project or have experience in a startup-like environment where you wore multiple hats.
3 spots only. Apply now.
How to Apply
Please submit your application using the form below.
We review applications on a rolling basis and encourage candidates from diverse backgrounds to apply. Subduxion is an equal opportunity employer committed to inclusive and fair hiring.