As a Senior Data Scientist focused on Computer Vision for autonomous UAVs, you will be a key contributor to the design, training, and deployment of on-board perception systems that enable safe, reliable flight. You'll specialize in real-time vision for tasks such as obstacle detection/avoidance, target recognition, landing, and mapping, and you'll work with large, diverse image/video datasets - including simulated data - to deliver production - grade models that run under tight SWaP (size/weight/power) constraints. You'll collaborate closely with autonomy, controls, firmware, and safety/compliance teams to ship a world - class perception stack that meets operational and regulatory requirements across multiple markets.
Key Responsibilities:
Model Development & Deployment
- Design, develop, train, evaluate, and deploy computer vision models for autonomous UAV applications, including: Object Detection/Tracking, Semantic/Instance Segmentation, Depth & 3D Perception, Multi-Sensor Fusion, Image/Signal Quality Assessment.
- Work with large, heterogeneous datasets (real flight logs, edge cases, nighttime/thermal, adverse weather) plus simulation and synthetic data with domain randomization.
- Implement data preprocessing, augmentation, active learning, and scalable labeling strategies.
- Use PyTorch/Tensor Flow and OpenCV; productionize models with ONNX/TensorRT, quantization/pruning, and other inference optimizations for edge accelerators (e.g., NVIDIA Jetson).
- Integrate models into the autonomy stack (ROS 2, PX4/ArduPilot) with real-time interfaces, fault detection, and safe fallback behaviors.
- Deploy and monitor models in production; build telemetry loops for continuous improvement.
Data Analysis & Insights
- Define and track performance metrics (e.g., mAP, MOTA, latency/FPS, power/thermal budgets) and go/no-go criteria for flight.
- Analyze failure modes from flight tests and post-flight logs; deliver clear recommendations to product/engineering.
Collaboration
- Partner with product and mission teams to translate requirements (range, altitude, environment, targets) into CV solutions.
- Work closely with autonomy/controls/firmware to ensure end-to-end system performance and timing guarantees.
- Collaborate with safety & compliance to align with airspace and operational regulations in target markets (e.g., Vietnam, Thailand, UAE), ensuring appropriate data handling and operational procedures.
- Share best practices with fellow data scientists; contribute to standards, code reviews, and documentation.
Research & Innovation
- Stay current with CV/robotics literature (Transformers for vision, event-based vision, uncertainty/OOD detection).
- Explore novel techniques (self-supervised learning, NeRF/occupancy mapping, learned sensor fusion) and evaluate them in SITL/HITL and flight tests.
- Champion MLOps: experiment tracking, model registry, CI/CD for ML, reproducible training, and model monitoring on-drone and in the cloud.
Qualifications:
- Bachelor's degree, Master's or PhD in Computer Science, Robotics, EE, or related field (or equivalent experience).
- 5–8+ years in data science/ML with a strong focus on computer vision and at least 2+ years deploying models to real-time or safety-critical systems.
- Proficiency in Python (preferred) and C++ for performance-critical on-board code.
- Deep knowledge of modern CV: detection/tracking, segmentation, depth/optical flow, VIO/SLAM, 3D geometry (camera models, epipolar geometry, PnP).
- Hands-on with ROS/ROS 2 and flight stacks (PX4/ArduPilot), including time synchronization and message pipelines.
- Experience deploying to edge hardware (e.g., NVIDIA Jetson Orin/Xavier) with TensorRT/ONNX; comfort with quantization/pruning and memory/latency trade-offs.
- Strong sensor calibration experience (mono/stereo cameras, IMU, LiDAR) and fusion (EKF/UKF or learned).
- Proven track record of production deployments, telemetry-driven iteration, and operating within latency, power, and thermal constraints.
- Excellent communication; ability to lead/mentor and collaborate across autonomy, firmware, and product.
Preferred:
- Prior work on autonomous robots/UAVs or other safety-critical platforms.
- Experience with simulation (AirSim, Gazebo/Ignition), synthetic data, and domain randomization.
- Familiarity with thermal/event cameras, nighttime/low-light perception, and adverse-weather robustness.
- Knowledge of mapping (TSDF/occupancy grids, loop closure), NeRF/implicit fields, and multi-view geometry at scale.
- Exposure to cloud platforms (AWS/GCP/Azure) for data pipelines and model training; Kubernetes is a plus.
- Experience working under regional regulations/operational constraints (e.g., Southeast Asia/Middle East operations).
Skills:
- Exceptional technical skills in computer vision and machine learning; strong analytical/problem-solving abilities.
- Clear, concise communication and cross-functional collaboration.
- Ability to own problems end-to-end, operate independently, and drive measurable outcomes.
- Passion for building innovative, high-impact perception solutions that make autonomous flight safer and more capable.
