Member of Scientific Staff, Computational Physics & Scientific Evaluation

Company DescriptionWe are a stealth AI-for-science company. Our mission is to accelerate scientific discovery by deploying AI that learns from real-world experiments. We are a small, technical team operating across India and the US. We are backed by exceptional angels and advised by industry leaders and researchers from OpenAI, Google DeepMind, xAI, Cursor, Hugging Face, etc.Role DescriptionYou will develop the datasets, benchmarks, and evaluation systems that tell us whether AI models can do useful scientific work. Working closely with the founding team, you will translate real physics and engineering problems into reproducible evaluation environments with well-defined inputs, ground truth, baselines, metrics, and pass conditions. Your work will connect numerical simulation, scientific machine learning, AI agents, and experimental evidence.This is a hands-on research and engineering role. You will work on hard problems, run simulations, inspect data, reproduce results, challenge model claims, and help determine what we build next.ResponsibilitiesScientific Evals and BenchmarksDesign evaluations for scientific models and agents across reasoning, simulation, prediction, tool use, and experiment selection.Turn real scientific workflows into bounded tasks with objective success criteria.Establish classical numerical, statistical, and machine-learning baselines.Measure physical consistency, predictive accuracy, calibration, robustness, out-of-distribution behavior, and computational efficiency.Evaluate whether models improve real decisions, such as selecting simulations or experiments.Create adversarial and failure cases involving incorrect units, boundary conditions, assumptions, solver settings, or physical constraints.Build reproducible evaluation harnesses, reports, and regression tests.Scientific DataSource, inspect, clean, and structure simulation and experimental datasets.Represent units, coordinates, geometries, meshes, boundary and initial conditions, protocols, uncertainty, provenance, and failed runs correctly.Define dataset schemas, quality checks, splits, documentation, and versioning.Identify leakage, inconsistent measurements, unsupported assumptions, and other issues that can invalidate an evaluation.Combine public data, partner data, simulations, literature, and experimental measurements where appropriate.Computational Physics and ModelingBuild and validate simulation workflows for physics-rich systems.Develop surrogate models, reduced-order models, and uncertainty-aware predictors.Compare learned models against numerical solvers, established analytical results, and physical measurements.Work on inverse problems, parameter estimation, optimization, and experiment selection.Help connect high-fidelity simulation with sparse experimental data and prospective physical validation.Contribute to technical decisions about model architectures, data requirements, and evaluation methodology.Research and CollaborationRead and reproduce relevant scientific and machine-learning research.Work with AI researchers, domain experts, laboratories, and industry partners.Translate loosely defined scientific questions into executable research plans.Communicate assumptions, limitations, negative results, and uncertainty clearly.Help establish a rigorous internal standard for evaluating scientific AI systems.QualificationsPhD strongly preferred in computational physics, applied mathematics, scientific computing, mechanical engineering, aerospace engineering, chemical engineering, materials science, or a closely related field. Exceptional candidates with a relevant master’s degree and substantial research experience are also encouraged to apply.Strong foundation in numerical methods, differential equations, statistics, optimization, and computational modeling.Experience building or validating simulations for physical systems.Strong experience with scientific computing tools.Experience working with scientific datasets, including quality control, uncertainty, provenance, and reproducibility.Ability to turn an open-ended research question into a rigorous experiment with clear baselines and metrics.Strong technical writing and communication skills.Willingness to work on-site in Bengaluru and collaborate closely with the founding team.The more the betterScientific machine learning, neural operators, physics-informed learning, surrogate modeling, or reduced-order modeling.Uncertainty quantification, Bayesian methods, active learning, design of experiments, or optimal experimental design.CFD, FEA, multiphysics simulation, molecular simulation, or other computational-science workflows.OpenFOAM, ANSYS, COMSOL, Abaqus, SU2, Fenics, or similar simulation tools.High-performance computing, GPU computing, or distributed data generation.Thermal systems, fluid mechanics, materials, electronic packaging, manufacturing, or other experimentally grounded domains.Designing evaluations for foundation models, LLMs, multimodal models, or AI agents.Using AI agents to accelerate literature review, code generation, data analysis, simulation setup, and research workflows while maintaining scientific verification.Experience working with physical experiments or collaborating with experimental researchers.Why JoinThe opportunity to define how scientific AI systems should be measured and trusted.A close working relationship with the founding team and significant influence over the technical direction of the company.Work spanning physics, machine learning, data, simulation, AI agents, and real-world experiments.The chance to help build a new scientific capability from India with global ambition.