ML Ops Engineer

EviSmart is a B2B SaaS platform transforming how dental labs, dentists, and design centers work together. We connect practices across 28+ countries with AI-powered workflow automation, overnight CAD design services, and intelligent case management. We're a team of ~60 across Vancouver, Manila, and Seoul—building the operating system for modern dentistry.

The Opportunity Our core AI research is done. The 3D segmentation models work. The generative models are built. The frontend is live, the storage layer is operational.

What's missing is the production backbone—the compute layer that turns research-grade models into fast, reliable, scalable services that real users depend on every day.

This role is about building that backbone. You'll architect and operate the infrastructure that powers heavy 3D mesh processing and GPU-intensive workloads for both internal teams and external customers. You'll own the systems that make our AI actually work in production—reproducibly, efficiently, and at scale.

This is a systems-level engineering role, not model research. If you're the person who bridges the gap between "it works in a notebook" and "it works for 10,000 users," we want to talk to you.

What You'll Be Doing

Convert research models into scalable, production-grade inference services

Design reproducible, version-controlled Docker environments with clear deployment standards

Optimize GPU memory usage and runtime performance for 3D workloads

Implement stateless service architectures with model versioning and rollback capabilities

Building Hybrid GPU Infrastructure

Architect asynchronous job orchestration across on-prem GPU servers and cloud GPU instances

Implement GPU scheduling, isolation, and intelligent workload routing

Design for concurrent users with fault tolerance and graceful failure recovery

Support hybrid infrastructure with cloud burst capacity for peak demand

Designing a Scalable Inference Platform

Build modular pipelines with clean separation between preprocessing, model inference, and postprocessing

Create well-defined service boundaries between backend APIs and compute services

Establish a platform architecture that allows future models to integrate without major refactoring

Deployment & Automation

Build CI/CD pipelines for model updates and infrastructure changes

Implement version control, rollback strategies, and release management across environments

Automate deployment workflows across hybrid infrastructure

Reliability & Observability

Implement logging, monitoring, and system observability across the compute layer

Track GPU utilization, job performance, and queue health

Design retry mechanisms and automated failure handling

Optimize latency and throughput to maintain stability under variable load

What We're Looking For Required

5+ years in ML Infrastructure, MLOps, or backend systems engineering

Proven track record deploying ML models into production environments

Strong expertise in Docker and container lifecycle management

Hands‑on experience managing and optimizing GPU‑based workloads

Experience designing asynchronous job systems or queue‑based architectures

Experience integrating ML services with backend APIs (FastAPI, Flask, etc.)

Preferred

Kubernetes or other container orchestration platforms

CI/CD pipeline design and automation

Monitoring and logging tools (Prometheus, Grafana, ELK, etc.)

Experience with high‑compute 3D ML workloads

Model versioning systems and experiment tracking

You’ll Thrive Here If You…

Have successfully transitioned ML systems from research to commercial deployment

Think in terms of platform architecture, not one‑off deployments

Care about reliability, scalability, and production SLAs

Are comfortable debugging across GPU, container, and infrastructure layers

Design systems for long‑term extensibility—not just what works today

#J-18808-Ljbffr