Will Johnson

Software Engineer · AI & Systems · IBM Research

About

I’m a Software Engineer focused on foundation model training, ML systems, and high-performance pipelines. I work at IBM Research on their AI Platform Engineering team where I build model tuning infrastructure, design lightweight transformer variants for accelerated testing, and develop CI/CD pipelines for AI hardware systems. I’m also building a game with friends in my free time.

Projects

AIU Spyre CI/CD Pipelines

Jenkins CIKubernetesPythonDevOps

Built scalable Jenkins pipelines validating end-to-end hardware/software compatibility for IBM’s AIU hardware accelerator, Spyre. Automated hardware-aware testing and artifact propagation across clusters, environments, and driver/library configurations, reducing manual intervention during release qualification.

Lightweight Transformer Model Variants

PythonPyTorchTorchTitanDistributed Training

Designed and trained four compact transformer models based on priority architectures using TorchTitan to accelerate nightly software stack regression testing on IBM’s AIU hardware accelerator. Reduced wall-clock test time by up to 90%. Released one model publicly (micro-g3.3-8b-instruct-1b), achieving 100,000+ Hugging Face downloads.

IBM Research Foundation Model Tuning Stack

PythonPyTorchTransformersCI/CDModel Fine-Tuning

Maintainer and core contributor for IBM’s open-source tuning stack used across IBM Research, Red Hat OpenShift AI, and IBM watsonx.ai. Increased CI test coverage 4x while reducing release cycle time 3x. Designed a flexible, multi-format data preprocessor enabling streamlined model tuning pipelines.

Hermitage Games

GodotGDScriptGame ArchitectureTeam Leadership

Co-founded an indie game company building a 2D action platformer in Godot. Leading a 10+ person cross-disciplinary team spanning engineering, art, narrative, and sound. Setting direction and timeline while contributing gameplay architecture, animation tooling, and in-engine implementation.

Skills

Languages

PythonC++Cx86/ARM AssemblyGDScript

ML Frameworks

PyTorchvLLMTensorFlowTorchTitan

Modeling

TransformersPretrainingLoRA TuningEval & Inference Optimization

DevOps & Systems

KubernetesJenkins CICluster SchedulingDistributed Training

Software

GitCompiler OptimizationData Structures & Algorithms