The Opportunity: Build AI That Understands Action

We are looking for a versatile and hands-on Computer Vision Engineer to be a cornerstone of our technical team. This is your chance to go beyond static object detection and work on the truly challenging problems in video AI. You will be responsible for building systems that can understand actions, track interactions over time, and identify complex events as they unfold.

You are the perfect candidate if you get excited about building a system that can tell the difference between a person walking past a machine and a person operating it. You'll be a critical voice in our product's development, with direct impact, visibility, and ownership.

Key Responsibilities

• Develop Video Understanding Models: Go beyond simple detection. Implement, train, and fine-tune sophisticated models for action recognition, activity detection, and temporal analysis. The goal is to build a system that understands context and behavior over time.
• Architect End-to-End Systems: Design and build robust backend systems and APIs (primarily in Python/Flask/FastAPI) to serve these complex video understanding models.
• Optimize for Real-Time Performance: Engineer high-performance video processing pipelines. You will work extensively with tools like NVIDIA DeepStream or similar SDKs to deploy models that can analyze streams in real-time.
• Manage the ML Lifecycle (MLOps): Take ownership of the entire model pipeline, using tools for data versioning (like DVC), experiment tracking (like MLFlow), and continuous deployment.
• Deploy to the Edge and Cloud: Containerize your applications using Docker and deploy them across a range of environments, from powerful cloud servers to resource-constrained edge devices.

Qualifications: What We're Looking For

We're seeking a practical builder with a strong engineering foundation and a passion for solving the complex temporal puzzles inherent in video data.

Must-Have Skills & Experience:

• Professional Experience: 0-2 years of professional experience in a role focused on AI/ML and software engineering.
• Core CV & Video Understanding: Proven, hands-on experience with object detection/tracking (e.g., YOLO, DeepSORT) and a strong, demonstrable interest or practical experience in video understanding techniques (e.g., action recognition, temporal models).
• Python & Backend: Strong proficiency in Python and experience building backend APIs and services (e.g., using Flask, FastAPI).
• Deployment Experience: You have deployed machine learning models into a production environment. Direct experience with NVIDIA DeepStream/TensorRT is a massive plus.
• DevOps Mindset: You are comfortable with Docker for containerization and utilize Git for all your version control needs.

The Ideal Candidate Profile:

You have a deep curiosity for how to model time in video data.

You have worked on projects that involve processing live video streams (RTSP, etc.).

You understand the trade-offs between model complexity, accuracy, and inference speed, especially for video.

You are familiar with message queueing systems (e.g., RabbitMQ, Kafka) and their role in processing event streams.

You have a portfolio or GitHub profile showcasing projects that go beyond simple image classification or detection.

Bonus Points (Great to Have):

Experience with video-specific architectures like 3D CNNs or Video Transformers.

Familiarity with multi-modal approaches, combining vision with text (LLMs).

Experience with the full MLOps lifecycle (DVC, MLFlow)