Director, AIML & Scientific Computing Optimization

Director, AIML &  Scientific Computing Optimization
The Onyx Research Data Tech organization is GSK’s Research data ecosystem which has the capability to bring together, analyze, and power the exploration of data at scale. We partner with scientists across GSK to define and understand their challenges and develop tailored solutions that meet their needs. The goal is to ensure scientists have the right data and insights when they need it to give them a better starting point for and accelerate medical discovery. Ultimately, this helps us get ahead of disease in more predictive and powerful ways.
Onyx is a full-stack shop consisting of product and portfolio leadership, data engineering, infrastructure and DevOps, data / metadata / knowledge platforms, and AI/ML and analysis platforms, all geared toward:​
Building a next-generation, metadata- and automation-driven data experience for GSK’s scientists, engineers, and decision-makers, increasing productivity and reducing time spent on “data mechanics”.
Providing best-in-class AI/ML and data analysis environments to accelerate our predictive capabilities and attract top-tier talent.
Aggressively engineering our data at scale, as one unified asset, to unlock the value of our unique collection of data and predictions in real-time.

Our Compute/AIML Performance and Optimization Engineering team is focused on optimizing first-in-class Compute and AIML platforms that accelerate application development, scale up computational experiments, and integrate all computation with project metadata, logs, experiment configuration and performance tracking over abstractions that encompass Cloud and High-Performance Computing. This metadata-forward, CI/CD-driven platform represents and enables the entire application and analysis lifecycle including interactive development and explorations (notebooks), large-scale batch processing, observability and production application deployments. The optimization team’s focus is on maximizing scale and performance of all aspects of the platforms.
 
A Director of Compute/AIML Optimization Engineering is a deeply technical leader. They consistently deliver major compute and AIML platform features and solutions with cross-organizational impact and value. They are recognized as expert in software engineering, scientific computing, and applications with deep understanding of performance and optimization, within the Onyx team, across R&D Digital & Tech, and even externally. They can work closely with -- and have strong technical knowledge of – underlying platform dependencies such as DevOps, Infrastructure and Cloud and can enable collaborations and help drive the requirements across other Onyx engineering teams that results in improved performance and better user experience. 

This role is responsible for building and leading a team of world-class software engineers focused on optimizing a best-in-class Compute and AIML Platforms at scale, quality, and cost. The Director of Compute/AIML Optimization Engineering will support the Sr. Director of Computing, Analysis, and AI/ML Platforms in building a strong culture of accountability and ownership in their team, as well as instilling best-in-class engineering practices (e.g. testing, code reviews, DevOps-forward ways of working). They work in close partnership with AI/ML Platform team, Compute Platform team, Data and Infrastructure Engineering, Product Management, Portfolio Management, and other engineering functions to ensure close alignment with customers and with engineering teams both upstream and downstream of their work.

Key Responsibilities:

Build, lead, develop, and retain world-class software engineers
Serve as a top expert for the optimization team, and contribute technical expertise to teams in closely aligned technical areas such as DevOps, Cloud and Infrastructure
Lead design of major optimization software components of the Compute and AIML Platforms, contribute to development of production code and participate in both design reviews and PR reviews
Accountable for delivery of scalable solutions to the Compute and AIML Platforms that supports the entire application lifecycle (interactive development and explorations/analysis, scalable batch processing, application deployment) with particular focus on performance at scale
Partner with both AIML/Compute platform teams as well as scientific users to help optimize and scale scientific workflows by utilizing deep understanding of both software as well as underlying infrastructure (networking, storage, GPU architectures, …)
Direct scrum team leads, and contribute technical expertise to teams in closely aligned technical areas 
Able to design innovative strategy and way of working to create a better environment for the end users, and able to construct a coordinated, stepwise plan to bring others along with the change curve
Standard bearer for proper ways of working and engineering discipline, including CI/CD best practices and proactively spearhead improvement within their engineering area
Serve as a technical thought leader and champion: e.g., speak at industry events, promote GSK as an attractive place to build a career and thrive as a Platform engineer, act as a key knowledge holder for the Onyx organization.

Why You?
Basic Qualifications:

Bachelor’s, Master’s or PhD degree in Computer Science, Software Engineering, or related discipline.
 8+ years of experience using specialized knowledge in cloud computing, scalable parallel computing paradigms, software engineering, CI/CD with Bachelor’s.
6+ years of experience using specialized knowledge in cloud computing, scalable parallel computing paradigms, software engineering, CI/CD with Master’s. 
4+ years of experience using specialized knowledge in cloud computing, scalable parallel computing paradigms, software engineering, CI/CD with a PhD. 
At least 2 years of experience with recruiting, managing, and developing engineers or other deeply technical contributors

Preferred Qualifications:

Deep experience using at least one interpreted and one compiled common industry programming language: e.g., Python, C/C++, Scala, Java, including toolchains for documentation, testing, and operations / observability
Hands-on experience with application performance tuning and optimization, including in parallel and distributed computing paradigms and communication libraries such as MPI, OpenMP, Gloo, including deep understanding of the underlying systems (hardware, networks, storage) and their impact on application performance
Expert understanding of AIML training optimization, including distributed multi-node training best practices and associated tools and libraries as well as hands-on practical experience in accelerating training jobs
Understanding of ML model deployment strategies, including agent systems as well as scalable LLM model inference systems deployed in multi-GPU, multi-node environments.
Deep expertise in modern software development tools / ways of working (e.g. git/GitHub, DevOps tools, metrics / monitoring, …)
Cloud experience (e.g., AWS, Google Cloud, Azure), including infrastructure-as-code and relevant tools and libraries such as Terraform, Ansible, and Packer
Experience with CI/CD implementations using git and a common CI/CD stack (e.g., Azure DevOps, CloudBuild, Jenkins, CircleCI, GitLab)
Experience with Docker, Kubernetes, and the larger CNCF ecosystem including experience with application deployment tools such as Helm
Experience with low level application builds tools (make, CMake) and understanding of optimization at the build and compile level
Demonstrated excellence with agile software development environments using tools like Jira and Confluence 
Deep familiarity with the tools, techniques, optimizations in high-performance applications space, including engagement with the open-source community (and potentially making contributions to such tools)
Experience with establishing software engineering ways of working and best practices for a team (whether informally or as formal SOPs etc)
Experience recruiting top engineering talent
Experience with agile planning and execution processes for software delivery