AI / Computer Vision Engineer

We are seeking a skilled AI / Computer Vision Engineer to join our team full-time in an on-site capacity. The ideal candidate will be responsible for designing, training, fine-tuning, and deploying advanced vision and vision-language models. This role involves working closely with foundational computer vision architectures, optimizing real-time inference pipelines, and managing deployments on edge and GPU platforms. Experience with human-in-the-loop workflows and active learning pipelines will be highly valued.

Key Responsibilities

- Model Training & Fine-Tuning: Develop and customize foundational vision and vision-language models such as YOLO, DINO, SAM, and various VLMs to meet specific domain requirements.
- Pipeline Optimization & Real-Time Inference: Enhance vision pipelines to ensure high-throughput, low-latency performance suitable for real-time streaming applications.
- Model Deployment: Convert, quantize, and deploy models using runtime frameworks like ONNX Runtime, TensorRT, and Triton on GPU and edge devices.
- Data & Active Learning: Implement human-in-the-loop processes and active learning pipelines to continuously improve dataset quality and model accuracy.
- Integration: Collaborate with engineering teams to embed computer vision modules into backend services and production environments.

Required Qualifications

- Core Architectures: Demonstrated hands-on experience with:
- YOLO (versions 8 through 11) for object detection and instance segmentation
- DINO / DINOv2 self-supervised vision transformers
- SAM / SAM 2 (Segment Anything Model) for interactive and promptable segmentation
- Vision-Language Models such as CLIP, Qwen-VL, LLaVA, and Florence-2
- Proven expertise in the machine learning lifecycle, including model training, fine-tuning, loss function optimization, and managing dataset edge cases
- Deployment & Inference: Proficiency in exporting and optimizing models using ONNX, TensorRT, or OpenVINO frameworks. Strong experience with CUDA and GPU acceleration, particularly for real-time inference constraints.
- Frameworks & Tooling: Advanced skills in PyTorch, OpenCV, and modern annotation or dataset management platforms.

Preferred Qualifications

- Human-in-the-Loop (HITL): Experience designing feedback loops where human annotators review uncertain model predictions to iteratively enhance training.
- Active Learning: Familiarity with automated data selection techniques such as uncertainty sampling and core-set selection to improve annotation efficiency.
- Edge AI Deployment: Experience deploying models on edge devices like NVIDIA Jetson or embedded Linux systems.

This position requires working on-site and offers the opportunity to contribute to cutting-edge AI and computer vision projects within a collaborative environment.