We are opening 2 remote internship position to join our team at the forefront of open research in NeuroAI. Our work explores the intersection of artificial intelligence and brain-computer interfaces (BCIs), using both invasive and non-invasive neural recordings.
As part of Tether Evo, you will contribute to high-impact, open-source projects focused on encoding and decoding brain activity leveraging AI.
Our mission is to advance scientific understanding while building real-world applications for neural interfaces, guided by transparency, accessibility, and collaboration.
Projects:
1) Compositional Structure in Speech-Related Neural Signals (6 months) — We are seeking an intern to investigate whether speech-related neural activity exhibits a compositional structure, where smaller reusable units such as phonemes combine to form larger units such as words. Recent work by Narasimha et al. in intracortical BCI has shown that neural activity underlying attempted handwriting contains reusable motif-like structure and that explicitly modeling this structure can improve decoding generalization and sample efficiency, particularly in low-data settings. Building on this idea, this project will ask whether an analogous organization exists in speech-related neural signals. A central challenge is that the available speech labels do not provide precise word- or phoneme-level timestamps, so the first phase of the project will focus on developing methods to segment continuous neural recordings into approximate intervals corresponding to individual speech units. Once such alignments are established, the project will analyze whether shared phonemic content across different words is reflected in shared neural structure—for example, whether the neural activity corresponding to the “book” portion of NOTEBOOK and COOKBOOK is more similar than the activity corresponding to their non-overlapping parts.
2) Bidirectional Modeling of Cursor Control from Human Intracortical Data (3 months) — This project will develop models for both decoding and simulating cursor-control neural activity from human intracortical recordings. Using publicly available datasets from Karpowicz et al. and Wilson et al., the first phase will focus on building a neural decoder that translates neural activity into cursor kinematics, including position and velocity. After achieving reliable decoding performance, the second phase will develop a neural encoder that predicts plausible neural activity from an intended cursor trajectory. The final goal is to integrate both models into a demonstration application that can simulate end-to-end cursor control, using either recorded neural data or synthetic neural activity generated by the encoder, and thereby provide a practical testbed for studying cursor BCI behavior in both real and simulated settings.
What We’re Looking For:
Candidates doing a PhD in computational neuroscience, AI, computer science, mathematics or similar fields, towards the end of their program and that could enroll in a 3-6 months industry internship. Also exceptional MSc or people that finished their PhD in the last year are eligible for the positions.
Strong proficiency in Python and PyTorch. Experience with object-oriented programming and knowledge of software engineering best practices for machine learning research (version control, testing, experiment tracking, reproducibility, and modular code design). Knowledge of HuggingFace transformers and diffusers interfaces.
Ability to efficiently implement methods from scientific papers
Research experience with invasive neural data or similar such as fMRI, EEG, MEG, ECoG
Excellent scientific communication and collaboration skills, publications in top-tier scientific journals or AI/neuroscience conferences are welcome.
If you’re passionate about advancing neuroscience and AI through open, ethical, and impactful research, we’d love to hear from you.
Positions are fully remote and applications are welcome from all over the world.