Computer Vision Engineer

About the role

Work at Blissway: Opportunity for Impact Every Day 🚀✨

Blissway is a startup that simplifies toll collection and dramatically improves highway safety. We are multiple startups in one: Deep Tech, AI/ML, Hardware, SaaS, and IoT. For the past five years, we have built a nearly insurmountable technological lead in tolling, a roughly $200B global industry expected to double in the next decade. We work for toll authorities and state DOTs, and we own the whole system, from the roadside camera to the final bill, so every vehicle we see becomes one accurate toll transaction and nothing more. While our competitors have thousands of employees, we operate with a lean but growing team of less than 30. Our impact is visible on the massive Interstate Highway System connecting every major US metro (except Juneau, AK—sorry, Juneau).

You are a good fit if...

- You love the grind: You take ownership and put in the time to meet deadlines. Our recent team survey showed an average of 55 hours/week, with occasional 70+ hour bursts for major releases.

- You are detail obsessed: You have experience writing code, designing or building systems that stands up to the unpredictability of the physical world, where the small details are the difference between success and failure.

- You are adaptable: We are a lean team. If you only want to work on a "niche thing" or are uncomfortable helping other teams when they need a boost, this isn't the place for you. We expect you to figure out what you need to get stuff done.

- You code really well: We mostly use Python and TypeScript, but we don’t care what you’re most proficient in today—as long as you learn fast.

- You value working together in person: You’re excited to work alongside the team at our Denver office. This role is in-office on Mondays, Wednesdays, and Fridays, with flexibility to work remotely on Tuesdays and Thursdays.

The Mission: Every image on our system has one job: complete an accurate toll transaction through our billing system. Your job is making that transaction more accurate. We process 11 million images a day, using detection, segmentation, classification, embeddings, and re-identification to match each vehicle to its toll. You'll find where our models fall short, decide what to fix first, and build the fix, from dataset curation to custom state-of-the-art architectures. This role is for the engineer who reads a new paper and immediately wants to know if it works in the real world.

The Work

- Own the Model: Data collection, dataset curation, design, training, and evaluation. Your job is making sure the model is correct, working closely with our MLOps Engineer to get it to production.

- Vision at Real Scale: 11 million images a day across vehicles, license plates, axles (so we classify each vehicle correctly), and lane markings. At this volume, the right model can drastically improve accuracy and cut cost at the same time.

- Classical CV to Custom SOTA: Our toolbox spans traditional computer vision algorithms through custom-trained state-of-the-art models. You pick the right tool, and if nothing off the shelf is good enough, you design and train your own.