About the job
About Us
Harmattan AI is at the forefront of innovation in defense technology, developing autonomous and scalable systems that enhance mission capabilities. Following our successful $200 million Series B funding round, which valued the company at $1.4 billion, we are actively expanding our teams to deliver critical systems to allied forces.
Our operations are driven by a commitment to creating technologies that make a significant impact, striving for excellence, setting ambitious objectives, and tackling challenging technical problems. In our fast-paced environment, we expect rigor, ownership, and execution from every team member.
About the Role
We are seeking a talented Software Engineer to join our Geometric Scene Understanding team in Paris. This role involves bridging the divide between advanced mathematical research and essential production software. While our Computer Vision (CV) engineers focus on developing core algorithms for 3D reconstruction—including SfM, SLAM, and Neural Radiance Fields—you will transform these intensive prototypes into high-performance, real-time engines designed for resource-constrained tactical hardware.
As the architect of our pipeline, you will process live drone telemetry and video, producing real-time 3D battlefield renderings that enhance the Common Operational Picture (COP) for our users.
Responsibilities
Pipeline Development: Design and implement a robust software pipeline to ingest, decode, and synchronize multiple drone video streams and telemetry data efficiently.
Performance Optimization: Analyze CV algorithms and optimize performance by leveraging GPU acceleration (CUDA, Vulkan, or TensorRT) and CPU multi-threading to achieve near real-time 3D reconstruction.
Spatial Data Management: Develop efficient 3D data structures, such as Octrees or Voxel grids, for storing, compressing, and streaming point clouds or meshes without exceeding memory limits.
Code Translation: Work collaboratively with CV engineers to convert Python/PyTorch prototypes into stable, production-ready C++ code that operates seamlessly for extended periods without memory issues.
Testing and Validation: Deploy builds on actual tactical PCs and evaluate performance metrics, including thermal management and GPU utilization during live drone operations.

