Probabilistic Programming Bayesian Jobs in California

Job Summary:
Sygaldry Technologies is building quantum-accelerated AI servers to exponentially speed up training and inference for AI. They are seeking a Research Scientist to define Quantum AI and work at the intersection of frontier AI/ML, quantum algorithms, and scientific machine learning.
Responsibilities:
• Develop and study models for high-dimensional scientific prediction, generation, and design, including:
• Diffusion models, flow matching, consistency models, score-based generative models, energy-based models, latent-variable models, autoregressive models, and normalizing flows.
• Scientific foundation models for molecules, materials, proteins, quantum systems, weather, climate, PDEs, and dynamical systems.
• Graph neural networks, geometric deep learning, equivariant models, neural operators, tensor methods, manifold learning, and learning on structured state spaces.
• Models that combine prediction, uncertainty, active learning, and closed-loop design for scientific discovery.
• Build algorithms and theory for the computational primitives that matter most for next-generation AI systems:
• Probabilistic inference, Bayesian modeling, variational inference, Monte Carlo methods, simulation-based inference, uncertainty quantification, and calibration.
• Optimization, sampling, amortized inference, sequential decision-making, Bayesian experimental design, reinforcement learning, planning, and control.
• Scientific reasoning systems, model-guided discovery, algorithmic discovery, and agents that can propose, test, and refine hypotheses.
• Benchmarking frameworks that reveal when a new computational substrate changes scaling behavior, not just constant factors.
• Identify where quantum computation can accelerate or reshape AI-relevant subroutines, including:
• Quantum algorithms for sampling, integration, Monte Carlo acceleration, linear algebra, optimization, Hamiltonian simulation, quantum simulation, and tensor-structured computation.
• Fault-tolerant quantum algorithms, resource estimation, complexity analysis, block encoding, QSVT, LCU methods, amplitude estimation, phase estimation, and quantum walks.
• Hybrid quantum-classical workflows where quantum primitives are embedded inside classical AI pipelines.
• New quantum-native model classes, kernels, embeddings, generative processes, and inference procedures that are mathematically motivated rather than benchmark-driven alone.
• Collaborate closely with quantum architecture, systems, and hardware teams to connect AI workloads to real machine requirements:
• Translate AI and scientific-computing bottlenecks into quantum resource requirements.
• Design benchmarks that compare quantum, classical, and hybrid approaches under realistic assumptions.
• Inform architecture choices by identifying the algorithms, error budgets, and primitives that matter for future AI workloads.
• Build prototypes in Python/JAX/PyTorch and, when useful, quantum software frameworks such as PennyLane, Qiskit, Cirq, CUDA-Q, TensorCircuit, or custom simulators.
Qualifications:
Required:
• Have a research record in machine learning, AI, statistics, physics, applied mathematics, computer science, quantum information, or a related field.
• Have deep expertise in at least two of the following: generative modeling, probabilistic inference, uncertainty quantification, geometric deep learning, graph neural networks, optimization, reinforcement learning/control, numerical methods, scientific machine learning, quantum algorithms, or quantum information.
• Have published research relevant to audiences at NeurIPS, ICML, ICLR, AISTATS, UAI, COLT, QIP, TQC, PRX Quantum, Nature, Science, or similar.
• Can move between theory and implementation: deriving algorithms, building prototypes, running careful experiments, and communicating results clearly.
• Are experienced with ML frameworks (PyTorch, JAX) and efficient inference implementation.
• Are excited to work with quantum hardware teams and help define what AI workloads should demand from future fault-tolerant quantum systems.
• Communicate complex ideas clearly across research communities.
• Value rigor: you are comfortable asking where quantum computation can help, where it cannot, and what evidence would distinguish the two.
Preferred:
• Research experience in diffusion/flow models, energy-based models, probabilistic programming, Bayesian deep learning, neural SDEs/ODEs, simulation-based inference, or scalable Monte Carlo.
• Experience with AI for science: molecular design, protein design, drug discovery, materials discovery, weather or climate prediction, quantum chemistry, PDE modeling, dynamical systems, robotics, or control.
• Experience with graph/geometric learning, equivariant architectures, neural operators, tensor networks, manifold methods, or structured world models.
• Background in quantum algorithms, computational complexity, quantum simulation, quantum chemistry, fault tolerance, resource estimation, or quantum information theory.
• Experience with JAX, PyTorch, CUDA/Triton, distributed training/inference, differentiable simulation, or high-performance scientific computing.
• A track record of publishing, open-source software, or building research systems that influenced a field.
Company:
Sygaldry Technologies is a tech company building quantum-accelerated AI servers to exponentially speed up AI training and inference. Founded in 2024, the company is headquartered in Mountain View, USA, with a team of 11-50 employees. The company is currently Early Stage.