We’re a small, funded startup building something genuinely different in machine learning - starting from physics first principles rather than iterating on transformers. I’m posting here because we specifically need someone with serious tensor network expertise.
The Technical Problem
We’re encoding classical feature spaces into quantum-inspired state representations using density matrices and tensor network structures. Think: taking ideas from DMRG/MPS methods and making them work for learning problems at scale, not just quantum simulation.
This isn’t “quantum ML” in the usual sense. We’re using tensor network mathematics because the problem structure demands it, not because it’s trendy. The challenge is making these methods actually work in production - which means handling numerical stability, contraction strategies, bond dimension optimisation, and all the things that make tensor networks hard.
What You’d Actually Be Doing
- Implementing and optimising tensor network contractions for our learning algorithms
- Deriving and testing different ansätze for representing our state spaces
- Debugging why theoretically sound approaches break at scale (the usual: numerical issues, contraction order, truncation errors)
- Moving between mathematical derivations and production code daily
- Working with reinforcement learning on top of these representations
We Need Someone Who
- Has a physics PhD and genuinely understands tensor networks (not just used a library once)
- Has worked with DMRG, MPS, PEPS, or related methods - ideally in research or industry
- Can write production-quality code, not just research scripts
- Knows quantum mechanics well (we’re encoding classical problems into quantum state formalisms)
- Is comfortable with the experimental “try things and see what works” nature of early-stage technical work
Particularly Valuable (But Not Required)
- Experience with TeNPy, ITensor, or similar libraries
- Background in quantum information theory
- Previous work on using tensor networks outside traditional physics applications
- Understanding of computational complexity and numerical optimisation
Practical Details
- Fully remote, Europe/UK based (CET working hours)
- Small team (you’d be working directly with founders who have deep technical backgrounds)
- Early stage, so meaningful equity
- Stealth mode currently, but we can discuss details in conversations
Why I’m Posting Here
We’ve tried traditional ML hiring channels and keep getting people who think tensor networks means “reshaping PyTorch tensors”. The ITensor community actually understands what we’re building and why these methods matter.
Not a Good Fit If
- You want a well-defined research problem with a clear publication path
- You need complete specifications before starting work
- You’re only interested in proving theorems, not building systems
- Current deep learning approaches seem sufficient to you
To Apply
Email CV to: niv.kazdan@huberta.io
Happy to answer technical questions here or via email. We’re reviewing applications continuously and responding quickly to strong candidates.