Machine Learning Engineer
About the role
We are looking for an AI/ML Engineer to join our AI & Data Unit and work across two connected areas:
Computer vision and satellite imagery analysis: object detection, segmentation, multi-modal fusion and geospatial pipelines.
LLM-powered automation: agentic systems, RAG and analytical report generation.
You should bring hands-on expertise in either Computer Vision or LLM systems, practical machine learning experience, and a willingness to work across both areas. Basic Computer Vision knowledge is required regardless of your main specialization.
An understanding of satellite imagery and its characteristics is expected, but direct experience is not mandatory. If you have not worked with satellite imagery, relevant experience in Data Science or NLP can provide a foundation.
Many of our tasks connect the two areas: turning detection results into LLM-generated analytical reports, using agents to orchestrate imagery processing pipelines, and building natural-language interfaces for geospatial data.
Responsibilities
Computer vision and remote sensing
Build, fine-tune, optimize and deploy object detection and segmentation models for optical, SAR and other satellite imagery.
Research and develop multi-modal architectures combining SAR, optical, infrared, panchromatic imagery and time-series data.
Evaluate pretrained geospatial models and adapt them to company-specific tasks.
LLM and agentic systems
Design and implement multi-agent systems for data processing and analysis automation.
Build RAG pipelines and LLM-powered tools for internal and customer-facing products.
Develop evaluation frameworks to measure the quality and reliability of agentic systems.
Integrate LLM APIs, including Anthropic, OpenAI and DeepSeek, into production workflows.
Research and prototype agentic approaches, including worker agents, orchestrators and tool use.
Across both areas
Collaborate with technical and non-technical teams to translate practical needs into technical solutions.
Take solutions from research and prototyping to production.
Define and track metrics to evaluate model and system performance.
Core requirements
- 3+ years of hands-on experience in ML/AI engineering, with expertise in Computer Vision, Deep Learning or LLM systems.
- Practical experience developing, training or fine-tuning and evaluating machine learning models.
- 3-5 years of Python experience, including OOP and code that follows PEP guidelines.
- Experience with PyTorch, Docker and AWS services such as EC2 and S3.
Statistical analysis skills. - R&D experience and comfort with iterative experimentation, including prototypes that may not become production solutions.
- Basic knowledge of Computer Vision and an understanding of satellite imagery.
- Willingness to contribute to both Computer Vision and LLM projects.
Expertise in at least one of the following areas
Computer vision
- Object detection: DETR models such as RT-DETR or RF-DETR, YOLO, Faster R-CNN, or oriented object detection.
- Semantic and instance segmentation: U-Net, SAM, Mask R-CNN, Mask2Former or equivalent architectures.
- GPU computing and CUDA; experience with OpenCV and/or TensorFlow.
LLM and agentic systems
- Hands-on experience with LLM APIs and prompt engineering, including tool use, function calling and structured outputs.
- Experience with LangChain/LangGraph, GSDK, CrewAI or custom agent orchestrators.
- FastAPI, asynchronous Python and REST API design.
- RAG pipeline development: ingestion, chunking, retrieval, reranking, and prompt and cost optimization.
- Evaluation approaches such as LLM-as-judge, comparative scoring, RAGAS and human-in-the-loop evaluation.
- Audit logging and system observability.
Nice to have
- Geospatial pretrained models and tools: Prithvi, SatMAE, TorchGeo or SkySense.
- Multi-modal fusion techniques, including cross-attention and feature pyramid fusion.
- Experience with DOTA, xView, SpaceNet or SEN12MS datasets.
- Change detection, land use and land cover classification, or damage assessment pipelines.
- MCP or A2A integration experience.
- Agent safety approaches: risk classification, approval gates, immutable audit logs and guardrails.
- SQLite or graph-based models for managing agent state.
- Vector databases such as pgvector, Pinecone or Weaviate.
- AWS SageMaker, AWS Lambda for serverless inference, Vertex AI or QGIS.
What the Company Offers:
- Work on global products in the field of space technologies.
- Mentoring, colleague support, and an open culture without micromanagement.
- 20 working days of paid vacation.
- 100% paid sick leave
- Medical insurance.
How We Work
We value simplicity, efficiency, and initiative.
Every engineer contributes to products that reshape how the world observes Earth from space.