Senior ML Research Engineer – AI Gaming Tech Startup (Hybrid)

Role Description

LocusX is reimagining the game development pipeline by embedding intelligence at its core. Our

platform is the first AI-native platform for game bug fixing, connecting testers, developers, and

project leads in one intelligent workspace.

As a Senior ML Research Engineer, you'll prototype novel approaches to root cause analysis and

automated repair, then own them through to production. You'll work at the intersection of frontier

model research and production engineering, training and adapting open-source foundation

models — increasingly deployed as agentic systems — to reason over the full surface of game

development. Your work will span root cause analysis, duplicate and similarity detection,

performance and observability intelligence, commit attribution, and bug reproduction.

Role Description

• Build models that reason over source code, textualized binary representations, logs,

telemetry, images, video, and commit history to identify root causes of bugs

• Develop bug-inducing commit attribution — pinpointing the change that introduced a

defect or crash

• Prototype and ship agentic systems for root cause analysis and automated repair that

propose, validate, and iterate on hypotheses and candidate fixes

• Train and fine-tune frontier and open-source foundation models to reason over game-

domain artifacts — code, engine state, binary assets, scripts, and runtime telemetry —

with a bias toward generalization across game types, projects, engines, and codebases

• Research privacy-preserving and client-specific fine-tuning strategies that protect IP

while enabling continuous improvement

• Design multi-dimensional embedding and retrieval systems — across root cause,

subsystem, spatial, input, and performance contexts — for issue similarity and correlation

• Translate research prototypes into production code-intelligence services

Qualifications

• Have deep understanding of modern deep learning — transformers, embeddings, and at

least one of: code, multimodal, or graph models — with hands-on experience training

them at scale on large datasets

• Have strong ML research publications, or a track record of taking research prototypes to

production

• Are self-driven with strong research intuition and clear technical communication

• Operate effectively in a fast-paced research environment, and can scope and deliver

projects end-to-end

• Enjoy collaborating across research, engineering, and domain experts

Strong candidates may also have

• Experience adapting foundation models to specialized technical domains (code, scientific,

multimodal)

• Background in software defect analysis, automated program repair, or transferable

experience adapting general-purpose models to similar reasoning tasks

• Experience training RL agents to replay game sessions and reproduce bugs in games

• Federated learning or other privacy-preserving ML in production

• PEFT techniques (LoRA, QLoRA, adapters) for efficient fine-tuning

• RAG systems with vector databases

• Gaming or developer-tools background; familiarity with Perforce, JIRA, Unreal, or Unity