PhD Position (m/f/d): World Models for Software Architecture Synthesis (Constructor Fabric)
Core
PhD research on building a World Model for software architecture synthesis using machine learning and formal methods to predict consequences of architectural decisions.
Role type
PhD Researcher (Machine Learning & Formal Methods)
Builds
A formal architectural skeleton (GearSpec, AppGraph) and a World Model that predicts system behavior, costs, and failure modes for production software.
Domain
Software Engineering, Machine Learning, Formal Methods, Model-Driven Engineering
Deliverable
production ML models
Required skills
Probability and statistics, Linear algebra, Discrete mathematics and graph theory, Mathematical logic and formal methods, Sequential decision making
Preferred skills
Formal languages and semantics, Category theory, Causal inference, Queueing theory, Combinatorial optimisation, Representation learning theory
Technologies
Python, PyTorch Geometric, DGL, Z3, cvc5, TLA+, Lean, Coq, tree-sitter, Ray, Dask, W&B, MLflow, Hydra, JetBrains MPS
Responsibilities
Design formal composition algebras and DSLs for architectural skeletons, develop architecture synthesis algorithms using constraint solving and learned models, engineer local transition models and pipeline surrogates, reconstruct and generate large-scale software data pairs.
Seniority
PhD Candidate
