> Markdown version of [/jobs/ext/2729156-sr-computer-vision-engineer](https://www.wearedevelopers.com/jobs/ext/2729156-sr-computer-vision-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Sr. Computer Vision Engineer - **Company:** Conxai Technologies GmbH - **Location:** München, Germany - **Experience:** Expert - **Contract:** Permanent contract - **Skills:** 3d Models, Computer Vision, Graph Database, Software Engineering, Spatial Data Infrastructures, Pytorch, Large Language Models, Deep Learning, Information Technology, Data Analytics, Virtual Agents, Data Pipelines - **Published:** September 5, 2026 - **Apply:** https://startup.jobs/sr-computer-vision-engineer-3d-semantic-scene-understanding-conxai-technologies-gmbh-8288203 ## About the Role * MS / PhD in Computer Science, Robotics, Electrical Engineering or related field * 3+ years of industry experience in Computer Vision and Deep Learning * 2+ years of leading 3D Computer Vision projects, specifically, geometric deep learning, 3D reconstruction * Experience with physics engines, e.g., NVIDIA Isaac Gym, MuJoCo, PyBullet, etc. is a plus * Experience in Agentic AI implementations with GraphRAG, Langgraph/LlamaIndex is a plus * Exceptional implementation experience with Open3D / PyTorch 3D, reconstruction (multi-view stereo, surface reconstruction and mesh-fitting, e.g., with TSDF), 2D * 3D "lifting" * Thorough understanding of software design * Previous experience in a fast-paced technology startup environment is a plus * Fluent and articulate in English ## Description As a Senior ML Engineer, you will lead the development of the spatial reasoning engine for our agentic AI platform. Your work focuses on the intersection of 3D Semantic Reconstruction, Geometric Deep Learning, and Agentic Inference. You will be responsible for building pipelines that transform unstructured multi-modal data into structured, actionable Spatial Knowledge Graphs. You will prioritize topological accuracy and semantic grounding, over photorealistic neural rendering. You will design the logic that allows autonomous agents to navigate, reason about, and perform inference on complex 3D environments, ensuring that AI-driven insights are rooted in the physical and engineering constraints of the real world. What You'll Do * Semantic Scene Reconstruction: Develop algorithms for 3D scene representation that prioritize geometric primitives and semantic labels over pixel-accuracy. This includes surface reconstruction, occupancy mapping and volumetric segmentation * Multi-Modal Fusion: Architect systems that fuse panoptic segmentation representations from CONXAI's AEC Foundation model with 3D models to generate high-fidelity, labeled representations * Knowledge Graph Augmentation: Automate the augmentation of 3D spatial data to CONXAI's Spatio-Temporal Knowledge Graphs, from reconstructed 3D scenes, mapping the hierarchical and functional relationships between structural elements * Agentic Inference & Reasoning: Design agentic workflows that perform complex reasoning tasks directly on the STKG * Actionable Affordance Mapping: Implement methods to identify "affordances" within a 3D volume, defining how agents or users can interact with the environment based on its physical geometry and engineering logic * Optimization & Scaling: Deploy SOTA models, representations and inferred domain context into production use-cases that deliver significant value to customers ## Related Videos - [How Robots Learn to be Robots](https://www.wearedevelopers.com/videos/1632-how-robots-learn-to-be-robots) - [Getting Started with Machine Learning](https://www.wearedevelopers.com/videos/260-getting-started-with-machine-learning) - [Geometric deep learning for drug discovery](https://www.wearedevelopers.com/videos/264-geometric-deep-learning-for-drug-discovery) - [TresJS a new declarative ThreeJS as Vue components](https://www.wearedevelopers.com/videos/543-tresjs-a-new-declarative-threejs-as-vue-components) - [Photonic Computing: Programming a New Class of AI Accelerators (incl. Live Coding)](https://www.wearedevelopers.com/videos/100196-photonic-computing-programming-a-new-class-of-ai-accelerators-incl-live-coding) - [Debugging Machine Learning Code](https://www.wearedevelopers.com/videos/71-debugging-machine-learning-code) ## Related Articles - [Everything a Developer Needs to Know About MCP with Neo4j](https://www.wearedevelopers.com/magazine/604-everything-a-developer-needs-to-know-about-mcp-with-neo4j) - [The Biggest German Tech Companies](https://www.wearedevelopers.com/magazine/424-the-biggest-german-tech-companies) - [Graph and AI Trends 2026: Why Is AI Running but Not Yet Delivering?](https://www.wearedevelopers.com/magazine/680-graph-and-ai-trends-2026-why-is-ai-running-but-not-yet-delivering) - [Dev Digest 121 - AI goes offline](https://www.wearedevelopers.com/magazine/456-dev-digest-121-ai-goes-offline) - [Coffee with Developers - Maria Apazoglou - Making AI understandable for all in production](https://www.wearedevelopers.com/magazine/475-coffee-with-developers-maria-apazoglou-making-ai-understandable-for-all-in-production) - [Best US AI Conferences for CTOs in 2026: Build vs. Buy, Vendor Evaluation, and Peer Intelligence](https://www.wearedevelopers.com/magazine/736-best-us-ai-conferences-for-ctos-in-2026-build-vs-buy-vendor-evaluation-and-peer-intelligence)