Chile Based, Machine Learning Engineer

About this role

Blissway seeks a Machine Learning Engineer to own the complete pipeline for computer vision systems processing 11 million images daily across the US highway system. You'll build models that ship to production—from raw sensor data collection and dataset curation through deployment, monitoring, and real-world iteration on actual roadside hardware.

What you'll do

• Own end-to-end ML pipelines from raw sensor data to production inference, including dataset curation, training, deployment, and monitoring
• Develop and optimize computer vision models for detection, segmentation, classification, embeddings, and re-identification at scale
• Balance cloud and edge computing constraints, pushing inference onto roadside hardware with strict performance and power requirements
• Evaluate and select appropriate techniques ranging from classical CV algorithms to state-of-the-art deep learning models
• Collaborate across hardware, software, and infrastructure teams in a lean startup environment
• Test and iterate models on real devices deployed across highways

What they're looking for

• Machine learning and computer vision (2-6 years production experience)
• Python and TypeScript
• Model training, deployment, and monitoring in production
• Classical computer vision techniques
• Modern deep learning frameworks and architectures
• Edge computing and model optimization
• Software engineering best practices
• Data collection and dataset curation

Benefits

• Work on real-world impact visible across the US Interstate Highway System
• Direct access to owned hardware deployed in the field for testing and validation
• Lean team of ~30 with ownership and influence on technical decisions
• Exposure to deep tech, AI/ML, hardware, SaaS, and IoT across one startup
• Frontier ML work with access to latest research and techniques
• Processing and optimizing at massive scale (11M images daily)

Likely interview questions

• Walk us through a production ML project where you owned the entire pipeline from data collection to monitoring—what was the hardest part and how did you handle it?
• Tell us about a time you chose a classical computer vision approach over deep learning. Why was it the right call?