Pure neural models are powerful but opaque. Pure symbolic systems are explainable but brittle. We work at the intersection — building neuro-symbolic architectures where neural components handle perception and language, while symbolic components enforce structure, constraints, and auditability.
You should have a strong research or applied background in at least one of: differentiable programming, neural theorem proving, logic-augmented LLMs, probabilistic logic, or constrained generation. Familiarity with logic programming languages (Prolog, Datalog, Answer Set Programming) and the ability to integrate symbolic modules into PyTorch or JAX pipelines is important. Publications or open-source contributions in this space are valued but not required.
At Novnex we value intellectual rigour and the ability to move from research to working prototype. We work on real enterprise problems, not benchmarks.
We are recruiting at the senior level — minimum 4 years of relevant research or engineering experience in AI/ML, with demonstrated work at the neural-symbolic boundary.
