Member of Technical Staff - Foundation Model Architecture AI Infrastructure

📋 Description

• • At Vinci, we are at the forefront of building the essential operator intelligence infrastructure that powers modern hardware programs on a daily basis. We have already achieved a significant milestone by demonstrating that a single, unified foundation model can operate effectively out-of-the-box across diverse industries and complex production workloads. This model is currently trained on an extensive 45TB of structured physics data, capable of performing billion-voxel inference in production environments, and is already deployed within Tier-1 semiconductor and hardware ecosystems, operating across multiple physical scales and operator regimes. This is not a research prototype; it is production-grade infrastructure that is now poised for industrial-scale deployment.
• • Your role as a Member of Technical Staff - Foundation Model Architecture & AI Infrastructure will be pivotal in scaling our deployment to an industrial magnitude. This involves a strategic focus on increasing simulation throughput by two orders of magnitude, expanding our capabilities from billion-voxel to trillion-voxel domains, broadening operator coverage across nonlinear regimes, and supporting global, multi-entity deployments within Tier-1 ecosystems. Our overarching ambition is not to be a frontier AI lab, but to establish ourselves as the default operator intelligence layer that hardware companies rely upon.
• • The "Operator Frontier" is where our current unified model excels, operating across a subset of partial differential equations in real industrial settings. The next critical phase involves expanding this unified architecture to encompass a wider array of operators, including but not limited to Maxwell's equations, Elasticity, Plasticity, Navier Stokes, nonlinear constitutive systems, and coupled multiphysics interactions. Crucially, we are not pursuing separate models for each equation; instead, we are evolving a single, generalized operator foundation model that transcends industries, physical scales, and conditioning regimes, while simultaneously scaling in deployment volume.
• • In this role, you will own significant aspects of AI architecture and systems engineering, moving beyond low-level GPU kernel work to define and scale the core operator intelligence layer. Your responsibilities will include evolving the Foundation Architecture by designing and refining transformer variants tailored for structured spatial domains, exploring sparse and locality-aware attention mechanisms, building hierarchical attention across multi-resolution fields, and developing graph-transformer systems for multi-entity interactions. You will also focus on improving modeling depth across nonlinear operator regimes, embodying architectural ownership.
• • Furthermore, you will scale our Training & Continuous Learning capabilities by expanding distributed training beyond our current 45TB-scale datasets, improving generalization across heterogeneous operator distributions, designing scalable data and curriculum strategies, and maintaining reproducibility and determinism across distributed systems. Building robust feedback loops from deployed production environments will be key, ensuring the system grows in capability without fragmenting its design.
• • Architecting for Trillion-Scale Inference is another core responsibility. With billion-voxel inference already running, you will help design systems capable of scaling to trillion-voxel domains, effectively utilizing sparse and hierarchical computation, and meticulously balancing memory, compute, and communication. Maintaining production-grade stability and determinism, where throughput and reliability are paramount, will be essential.
• • Shipping at Industrial Scale means our models already operate within Tier-1 hardware programs. You will contribute to shipping expanded operator capabilities into production, increasing simulations per day by 100x, supporting global, multi-entity deployments, and maintaining robustness under diverse industrial workloads. Success will be measured by adoption, throughput, and reliability, rather than solely by leaderboard metrics.
• • We are looking for individuals with deep experience in large-scale foundation model architecture, including transformer variants (sparse, hierarchical, graph-based), distributed training systems, production ML system design, and scaling structured datasets. You should be adept at writing clean, maintainable, high-quality code and possess a mindset focused on architectural generalization, stability under nonlinear regimes, communication vs. computation tradeoffs, deterministic distributed execution, and designing systems that evolve into durable infrastructure. Your track record should include building AI systems that run in production, not just experiments.
• • Engineering expectations include strong software engineering fundamentals, proficiency in clean abstractions and scalable code design, experience with modern ML stacks like PyTorch and distributed training ecosystems, and a strong discipline in CI, regression testing, and validation. Comfort in evolving core model infrastructure is crucial, as this role is fundamentally about building lasting infrastructure.
• • At Vinci, you will benefit from a single model already deployed across industries, working with 45TB of structured training data, and contributing to billion-voxel inference in production. You will engage with Tier-1 customers operating on real hardware workflows, enjoy high ownership at the Series A stage, and have the opportunity to define a foundational abstraction layer early in our growth. We are building something that hardware companies will depend on daily. If you are driven to define and scale the operator intelligence layer that industry runs on, this role is tailor-made for you.