Data Scientist Lead

Purpose & Scope

As the Data Science Lead (coordinator), you will lead a team of data scientists and scientific software developers within our Rio de Janeiro R&D hub, working as one team with TGS Data Science in Houston. You will set the technical direction for developing, scaling and supporting TGS’s AI and data technologies: ML-driven processing and interpretation, the MDIO open-source data format and the TGS Data Verse platform — on GPU and CPU HPC clusters and in the cloud. Your primary objective is to turn research into robust, high-performance software that TGS and its clients rely on, in support of the TGS Data-AI strategy (AI for Growth).

Responsibilities

• Team Leadership: Lead, mentor and grow a team of data scientists and scientific software developers, setting clear technical goals, coding standards and development paths.
• Technical Direction: Define the architecture of ML models, data pipelines and HPC/cloud workflows in line with the TGS Data Science roadmap in Houston — one shared codebase and toolset.
• AI Development: Oversee the design, scaling and validation of deep learning models for seismic processing, imaging and interpretation, from seismic foundation models (ViT) to 3D CNNs such as SaltNet.
• Performance & Scalability: Drive the profiling, parallelization and optimization of training and inference on GPU and CPU clusters (CUDA, MPI, multi-node) and on AWS, keeping compute efficient and costs under control.
• Data Platforms: Champion MDIO and cloud-native data practices — chunked, compressed data accessible in place — and integration with TGS Data Verse (Data Lake, OSDU) and Prediktor operational data.
• Production Readiness: Own the path from prototype to product — packaging, testing, CI/CD, model registry and deployment into TGS production software (e.g., Imaging AnyWare) and HPC environments.
• Engineering Excellence: Establish best practices across research and production code: clean architecture, code review, automated testing, documentation and reproducible experiments.
• User & Business Support: Partner with geophysicists, imaging, HPC/IT and product teams to turn needs into requirements and to support users of TGS AI and data tools across the energy data value chain.
• Innovation & Representation: Track advances in AI, HPC and data technology; drive open-source contributions (e.g., MDIO), patents and publications; represent TGS in technical forums.
  
  

Education & Experience

• Education: Bachelor’s/Master’s degree in Geophysics, Physics, Computer Science, Applied Mathematics, Engineering or a related field, including 5+ years leading R&D or software projects.
• Leadership Experience: 3+ years leading data science, research or scientific software teams, with a track record of delivering software to production.
• Technical Experience:
• Expert Python (NumPy, SciPy, PyTorch) and modern C/C++; Fortran is a plus.
• Highly experienced with MPI, CUDA, OpenMP or other parallel programming models on GPU and CPU HPC architectures.
• Highly experienced in debugging, profiling and optimizing scientific code (e.g., Nsight, perf, VTune).
• Distributed training and inference of deep learning models (DDP/FSDP, mixed precision) and MLOps practices (experiment tracking, model registry, deployment).
• UNIX/Linux and POSIX programming, Git, automated testing and CI/CD.
• Data & Infrastructure: Multidimensional, chunked data formats (MDIO, Zarr, SEG‑Y) with Dask/xarray; AWS (S3, EC2 GPU); containers (Docker, Kubernetes); HPC schedulers (e.g., Slurm).
• Domain Expertise: Experience with seismic processing, imaging or interpretation algorithms; publications or patents in ML, HPC or geophysics are a strong advantage.
• Desirable: OSDU data platforms and master data management; industrial time-series data and OPC UA; contributions to open-source scientific software (e.g., MDIO).
• Languages: Fluent English and Portuguese
• Work arrangement & Location: Presential, based in Rio de Janeiro – RJ, with Brazilian residency and work authorization, and available for occasional travel.

Show more Show less