Apply on Employer Site

Parallel Domain · 2 months ago

Creative ML Technologist

Vancouver, British Columbia, Canada

Full-time

Hybrid

Entry, Mid Level

$120K/yr - $170K/yr

2+ years exp

Parallel Domain is focused on enabling the development of autonomous systems through their simulated testing platform. They are seeking a Creative ML Technologist to prototype and evaluate workflows that integrate generative AI with perception capabilities, ultimately enhancing the testing and deployment of AI technologies.

3D TechnologyAutonomous VehiclesBig DataMachine LearningSimulation

Responsibilities

Compose workflows: Build focused ComfyUI graphs that chain perception and diffusion steps to answer product questions

Integrate new models: Author/extend custom ComfyUI nodes in Python and create wrappers for new open-source models

Measure & compare: Produce side-by-side results with qualitative examples (paired frames) and simple, meaningful metrics to guide decisions

Scan & triage the model landscape: Track new diffusion and perception releases, curate a weekly shortlist, and convert promising ones into ComfyUI nodes for quick A/B tests

Document & hand off: Provide inputs, controls, expected artifacts, and failure modes with runnable .json graphs and enough version info for others to reproduce and extend

Collaborate & iterate: Partner with creative users to calibrate taste and with engineering to productionize proven workflows

Qualification

ComfyUIPythonDiffusion conditioningPerception tasksPyTorchEvaluation & communicationModel-scouting3D ReconstructionVFX/Games exposure

Required

Portfolio: 2–3 ComfyUI graphs you authored (.json + screenshots) and at least one custom node or wrapper

ComfyUI depth: Experience structuring multi-step graphs and keeping them readable and reusable

Diffusion & perception know-how: Practical use of diffusion conditioning (e.g., ControlNet/IP-Adapter), perception tasks (segmentation, matting, depth), and sampler/scheduler trade-offs

Python: Strong scripting and ability to write/extend ComfyUI nodes

PyTorch (basic): Can read and lightly modify model code when needed

Evaluation & communication: Clear write-ups of experiments and decisions with reproducible artifacts

Model-scouting habit: You stay current on new perception/diffusion models and can recommend/justify trials with an integration plan in ComfyUI