3 months ago
Tallinn, EstoniaMid Level
Responsibilities
- Train, fine-tune, and evaluate computer-vision models for object detection, image quality, and static-object or duplicate suppression.
- Own model serving with NVIDIA Triton ensembles, DALI preprocessing, TensorRT inference, post-processing, model repositories, and no-downtime reloads.
- Build and curate datasets through ingestion, labeling, quality control, and data-problem analysis using FiftyOne and Label Studio.
- Design regression suites, A/B comparisons, and evaluation metrics tied to real detection quality.
- Develop LLM/VLM and agentic capabilities using self-hosted model stacks, retrieval, and tool-using agents.
- Deploy and operate models reliably on real-world camera imagery across day/night conditions, IR/RGB inputs, weather, and poor connectivity.
Requirements
- Strong Python and modern ML experience, including PyTorch, model training and fine-tuning, and notebook-driven experimentation.
- Practical computer-vision experience with object detection and image data, including understanding real-world model failure modes.
- Experience taking models to production through model serving, optimization, and evaluation of live behavior against offline metrics.
- Ability to independently investigate ambiguous accuracy problems, run experiments, and ship measurable improvements.
- Rigorous understanding of datasets, ground truth, edge cases, and reliable evaluation.
- Two to five or more years of relevant ML engineering experience.
- Preferred experience with NVIDIA Triton Inference Server, TensorRT, DALI, or comparable GPU model-serving and optimization tools.
- Preferred experience with FiftyOne, Voxel51, Label Studio, LLM/VLM systems, vLLM, LoRA, RAG, multimodal models, tool-using agents, MCP, or LLM orchestration frameworks.
- Preferred MLOps experience with CometML, Opik, model registries, and reproducible training pipelines.
- Exposure to edge or IoT inference, resource-constrained systems, anomaly detection on device telemetry, or NATS/gRPC and other event-driven communication is beneficial.
Benefits
- Fast-growing business.
- Fantastic office in Tallinn.
- Down-to-earth and innovative company culture.
- Stebby wellness benefit.
- Additional vacation and health days.
Categories
