Forward Deployed Engineer (ML/Software)

Visia; New York (NY); Ibrido; USD 120.000 - 160.000; Tempo pieno

Vantaggi offerti da questo lavoro

Employer-paid health premiums

401(k) with company match

Equity

Unlimited PTO

Descrizione del lavoro

Visia in New York, NY is seeking a Deployment Engineer to take customer systems from initial site visits to a fully operational solution. You will manage cameras, edge compute, and site-specific models across industrial environments, with 30–50% on-site work and frequent travel to steel mills, scrap yards, and recycling plants.

You will collaborate with ML engineers, annotation teams, and Deployments Engineers to ensure high-quality production results, ship improvements, and build repeatable

Competenze

• 2–3 years of full-time experience in a related role.
• Hands-on experience shipping computer vision models used in production.
• Comfortable with frequent travel to plant sites and on-site deployments.

Mansioni

• Lead kickoff calls with customers to define success metrics.
• Bring up cameras and edge compute on site and deploy initial models.
• Fine-tune detection/segmentation models per site data.
• Coordinate with annotation team and review labeling quality.
• Package models for on-site GPUs and verify production consistency.
• Set up dashboards and alerts used by operators.
• Set realistic expectations and communicate progress and risks.
• Document tooling, checks, and playbooks to accelerate future deployments.

Conoscenze

Computer vision

Field deployments

Customer-facing

Edge compute on site

Descrizione del lavoro

Visia | New York, NY | Hybrid | Full-time

Base salary: $120,000–$160,000 per year

Take a customer from the first site visit all the way to a system their plant relies on every day. You'll own everything in between: the cameras and compute on site, the vision models, and the value the customer actually sees in Visia’s physical intelligence solutions.

The role

Visia builds Solstice, which combines cameras, infrared, X-ray and more with compute on site and models we train on each customer's own material. Solstice runs with hundreds of industrial operators across the US, mostly steel mills, scrap yards, and recycling plants. Customers use it to catch things like non-metallic contamination in a mill's scrap feed, copper and motors mixed into shredded steel, and pressurized gas tanks before they cause an explosion.

The hard part, and the reason this role exists, is that no two sites look the same. The material, the camera angles, what the customer counts as a good result all change from plant to plant, so a model that does well at one site almost always needs to be retrained on the next site's data before it holds up. You'll be the person doing that work and staying with the customer until they're getting the results they paid for.

You'll report to Brandon, who runs Deployments, and work day to day with our ML engineers, our internal annotation team, and the Deployments Engineers who maintain the hardware fleet. Expect to be on site around 30–50% of the time. Some trips are a week of data collection or commissioning, some are a day walking a plant manager through results, and some come up on short notice because a camera needs re-aiming or an enclosure needs rewiring.

What you'll do

• Agree on the target KPIs first. Run the kickoff calls and settle with the customer what success looks like before anything is installed.
• Get the site live. Bring up the cameras, edge compute, and first models, then fine-tune detection and segmentation models on the site's own frames.
• Improve the model with evidence. Go through the misses and false positives class by class, tune thresholds per camera, and ship a new model when the numbers show it's better.
• Work with our annotation team. Decide which frames to label and how, review their work, and feed back the hard examples the model keeps missing.
• Ensure quality. Package models for the GPUs we run on site and check that they behave the same in production as they did in testing.
• Own what operators see. Set up the dashboards, alerts, and laser marking the operators use, and make sure each one fires when it should.
• Manage customer expectations. Tell them what the system catches today, what it still misses, and when that will improve, and bring bad news early with a plan attached.
• Leave the next deployment easier. Write up what you fixed as tooling, checks, or playbooks so the whole team moves faster.

Must-haves

• At least two to three years of full-time experience. This role isn't a fit for new grads.
• Computer vision shipped in a real job. You've trained and shipped detection or segmentation models that a business depended on, rather than only class or side projects, and you can talk through how you measured them and where they fell short.
• Comfortable with frequent travel, including short-notice trips to steel mills, scrap yards, and recycling plants.

Nice-to-haves

• You like building and fixing physical things, and you're comfortable with a drill and a crimping tool on a plant floor.
• You've worked closely with a labeling team and seen how label quality moves a model's precision and recall.
• You've deployed models on NVIDIA Jetson or optimized them with TensorRT.

Expected impact

• 1 month in: You've learned our hardware and model pipeline, joined a teammate on a live deployment, and shipped a model improvement for one of our current customers.
• 3 months in: You're leading your own deployments and running the customer reviews, with the models, data, dashboards and alerts for each of your sites in your hands.
• 12 months in: You have a group of sites hitting their targets, and your playbooks have shortened how long it takes us to bring a new site live.

• 100% employer-paid premiums for employee coverage (medical, dental, vision)
• 401(k) with company match (100% of the first 3%, 50% of the next 2%)
• Meaningful equity with a 10-year post-termination exercise window
• Unlimited PTO

The range listed is base compensation. We offer participation in our bonus and variable compensation plans to select roles.