Aditya T.

Have an engineering simulation that runs, but is not yet trustworthy, repeatable, or ready for a design decision? I help teams turn existing CFD and multiphysics models into audited baselines, reproducible studies, ML-ready simulation data, and solver-validated optimization workflows. I am not tied to one solver. My delivered work spans OpenFOAM, CalculiX, a commercial cloud multiphysics platform, and client-provided or custom solver environments. I work at the layer that determines whether the complete workflow is dependable: • physical assumptions, boundary conditions, and source terms; • geometry, mesh, case configuration, and parameterization; • solver execution, stability checks, and numerical sensitivity; • automated case generation and simulation campaigns; • field extraction, VTK exports, manifests, and ML-ready data; • validation, supported claims, failure boundaries, and reproducible handover. Recent client work: • MultiFluidX (UK): inherited a custom OpenFOAM vegetation solver and reference case, compared three solver paths, corrected material physics and implementation issues, and built configuration-driven control-versus-intervention studies. Follow-on setup generated 48 cases across three façade sizes, accepted three meshes below two million cells, and verified the four required VTK fields through real smoke runs. Production remains in progress. • Quanscient (Finland): built surrogate-assisted design-space exploration for a commercial cloud multiphysics platform using Gaussian-process surrogates, feasibility-aware active learning, Pareto analysis, and solver validation. The client received execution-gated notebooks and an evidence pack its team could rerun. Public reference available. I can help with: • reproducing or auditing an inherited simulation workflow; • diagnosing boundary-condition, source-term, stability, or mesh issues; • converting manually edited cases into configuration-driven studies; • building and executing controlled simulation campaigns; • preparing traceable field data for downstream ML; • sensitivity screening and design-space reduction; • surrogate modeling and active learning when solver costs justify them. Typical handover: • editable cases, configuration, source changes, and run scripts; • case manifests, checksums, and export instructions; • numerical checks, plots, and engineering interpretation; • assumptions, limitations, failure modes, and next decisions. For a new proprietary package, I begin with a scoped bring-up milestone to confirm access, compatibility, and reproducibility before committing to production delivery. The solver remains the authority, and any AI layer must earn its place against independent solver evidence. Send me your current model, solver environment, latest log, and the decision the simulation needs to support. I can determine whether the right first milestone is a model audit, validated baseline, automated campaign, or solver-validated AI workflow.