Agentic AI Engineer (Computer Vision)

CVEDIA; United States; Remote; USD 120,000 - 180,000; Full time

Job summary

CVEDIA is hiring an engineer to build AI agent systems and guide neural network training for real-world applications. You will start by learning our automated model pipeline and then broaden into research, developing new models, and addressing diverse customer problems with agent-driven solutions.

The role emphasizes software engineering on Windows and Linux, with experience in running open-source models and staying current with AI agent techniques.

Qualifications

• Have trained object detectors or classifiers for computer vision.
• Own a GPU capable of running local models because you wanted to experiment.
• Have served models with engines such as vLLM or SGLang, or take part in their communities.
• Have a background or a strong interest in data science.

Responsibilities

• Learn, maintain, and improve our synthetic data pipeline and automated training system.
• Take over a model as your first project, and improve it with an agent loop that generates its own data as it goes.
• Design systems that handle false positives from the field, flagging only uncertain detections and fixing the rest automatically.
• Lead our research on vision-language models, fine-tuning them for specific tasks and using agents to track the latest techniques.

Skills

Agent harnesses

GPU for local models

vLLM / SGLang

Data science interest

Tools

Git

Windows

Linux

Job description

Описание:

CVEDIA · Remote, anywhere (with overlap with European afternoons) · Full-time (This job requires Advanced English (C1) proficiency - with the ability to communicate fluently and handle complex, fast-paced technical and professional tasks.) CVEDIA builds computer vision models and applications that run in real products on server and edge hardware around the world. We train our models on synthetic data, and much of our model work already runs through AI agents. We have made the full move to this way of working across the whole company, and we are looking for people who want to push it further. We are hiring an engineer who builds systems out of AI agents and knows enough about training neural networks to point those agents at the right problems. You are not just a software developer and not just a data scientist. You may have spent your evenings building your own agent harnesses, fine-tuning open-source models and running them on your own GPU, and getting groups of agents to finish real work. If so, you probably already know what this job is. You will start by learning our automated model pipeline, and then take on your first model. After that, the work widens to new research, new customer problems, and our software. You will design the harness, and the agents will do much of the work.

Задачи:

Learn, maintain, and improve our synthetic data pipeline and automated training system. Take over a model as your first project, and improve it with an agent loop that generates its own data as it goes. Build closed loops in which agents train detectors and classifiers, test them against validation sets, and improve them with little human input. Design systems that handle false positives from the field, flagging only uncertain detections and fixing the rest automatically. Lead our research on vision-language models, fine-tuning them for specific tasks and using agents to track the latest techniques. Solve new detection and classification problems for customers and partners by building new models or extending existing ones. Apply agents beyond models, for example to QA automation and to code bases that find and fix their own bugs. Share what works with the team, and present results to customers when a project needs it. You may be a good fit if You have built your own agent harnesses or agentic workflows, and you can show us the work. You have run many agents at once, and you have dealt with the usual problems: communication between agents, token use, and checking that tasks are really finished. You have a background in professional software engineering on both Windows and Linux, including Git and an issue tracker. You have fine-tuned or quantized open-source models, for example by training LoRAs, and you have run them locally. You know the model landscape from the early GPT models to today's open-source ones, including where each one is strong and where it fails. You communicate clearly, both within the team and with customers. You have a real passion for AI and agents, and you try new tools and ideas without being asked.

Требования:

• Have trained object detectors or classifiers for computer vision.
• Own a GPU capable of running local models because you wanted to experiment.
• Have served models with engines such as vLLM or SGLang, or take part in their communities.
• Have a background or a strong interest in data science.
• You do not need to tick every box. If you have the mindset, we will teach you the computer vision and research side.

Why CVEDIA Agents as your everyday tools. You will use Claude, Codex and self-hosted models every day, with no limits on applying them to your work. Work that ships. Our models already run in production on edge boxes around the world, and your systems decide how fast they improve. Real influence. We are a small team. You will work directly with all of us and help shape how the whole company works with AI. People who think like you. We discuss agent behaviour, harness design, and model quirks every day. You will have people to compare notes with.

Where and how you work

The role is fully remote. You need to overlap with European afternoons, which is when most of our meetings happen. Pay depends on your experience. We will discuss it with you early in the process.

CVEDIA welcomes applicants from every background and country. We judge your work, not your title or your path to it.