Full Stack Engineer - ML Tooling & Infrastructure - ON-SITE

Full-Stack Engineer, ML Tooling & Infrastructure

Location: Toronto GTA (on-site) 

Type: Full-time 

Openings: 1 

Salary: CA$130K – $180K + equity

The role

Be the force multiplier for a core computer vision and machine learning team. You won't build foundation models; you'll own the internal tooling, infrastructure and targeted ML services that let that team move fast. That includes the annotation platform the labeling team uses daily and a customer-facing privacy-blurring pipeline. This is a true hybrid role: practical ML deployment plus robust, user-facing full-stack apps.

What you'll do

• Own the annotation web app: new task types, review workflows, keyboard-driven UI and progress dashboards
• Write pipelines that move video frames, metadata and JSON between cloud storage, databases and the client app
• Build quality tools for inter-annotator agreement, audit sampling and dataset health
• Own the pipeline that detects and blurs faces, screens and license plates in delivered footage
• Deploy lightweight detection and segmentation models so annotators correct rather than create from scratch
• Optimize PyTorch models (TensorRT, ONNX) and serve them through fast, concurrent APIs

What you bring

• Heavy frontend engineering: interactive web apps in React, Vue or modern JS/TS, including complex state, canvas rendering or in-browser video
• Production-grade Python: fast, concurrent APIs (FastAPI, gRPC) that handle high-volume media
• Practical applied ML: fine-tuning and deploying detection, segmentation or tracking models on messy real-world data, beyond the notebook

Nice to have

• Built or customized an annotation UI (bounding boxes, polygons, keypoints) and understand human-in-the-loop pain points
• Video pipelines: FFmpeg, frame extraction, codecs, latency optimization
• Active learning or hard-negative mining to improve datasets
• You measure success by how fast your end users can work

Not the right fit if you want to design novel architectures, prefer backend-only work, or expect clean datasets handed to you.