Computer Vision Engineer

T3i Inc.•San Antonio, TX

3h•Onsite

About The Position

T3i builds mission-critical unmanned systems designed to improve U.S. forces' operational effectiveness and battlefield lethality. Our Blackfoot platform is a tactical FPV drone system engineered for reliability, adaptability, survivability, and rapid deployment in demanding operational environments. We operate in tight feedback loops between design, build, test, and field use, where every flight directly informs the next iteration of the platform. We are a small, fast-moving team that values ownership, technical depth, execution, and mission focus. We are seeking a Computer Vision Engineer (CVE) to develop the Blackfoot’s perception and visual autonomy stack. This role is critical to delivering robust onboard vision capabilities - object detection, tracking, visual-inertial navigation, and last-mile terminal guidance - that perform reliably in degraded, GPS-denied, and EW-contested environments. The CVE will own the design, training, optimization, and deployment of computer vision models running on resource-constrained embedded hardware, working directly with flight software, autonomy, and hardware teams to translate algorithms into deployed capability. The CVE will lead the development of real-time perception algorithms for the Blackfoot platform, including detection, classification, tracking, visual odometry, and target lock / terminal guidance. This role emphasizes building models and pipelines that are not only accurate, but small, fast, and resilient under real-world operating conditions - motion blur, low light, adverse weather, occlusion, and adversarial RF/visual environments. The ideal candidate is equally comfortable writing custom CV algorithms from scratch, training and optimizing neural networks, integrating models onto embedded computers (Jetson, Hailo, Ambarella, or similar), and iterating against real flight data captured by our test pilots. Strong classical computer vision fundamentals - camera calibration, multi-view geometry, feature matching, and bundle adjustment - are as important as modern deep learning. This role is ideal for someone who thrives in fast-paced R&D environments where models are deployed to airframes within days, not quarters, and where flight-test feedback directly drives the next training cycle.

Requirements

Bachelor's degree in Computer Science, Electrical Engineering, Robotics, Applied Math, or related field
5+ years of hands-on experience developing and deploying computer vision systems (3+ years with an MS, or 2+ years with a PhD)
Strong proficiency in modern C++ (C++17 or later) and Python, including writing high-performance code for real-time systems
Demonstrated experience writing custom computer vision algorithms - not solely applying existing libraries - optimized for real-time performance
Deep experience with PyTorch or TensorFlow, including custom training pipelines, loss design, and large-scale dataset workflows
Solid foundation in classical computer vision: camera calibration, multi-view geometry, feature detection/matching, optical flow, bundle adjustment, and pose estimation
Solid foundation in modern deep learning architectures for vision (CNNs, transformers, detection/segmentation/tracking heads)
Demonstrated experience deploying CV models to embedded or edge compute platforms with hard real-time and SWaP constraints
Working knowledge of model optimization techniques: TensorRT, ONNX, quantization (INT8/FP16), pruning, distillation, and compiler-level tuning
Experience with OpenCV, Eigen, Ceres, and standard CV/robotics tooling
Experience building and maintaining data pipelines for image and video data at scale
Ability to validate perception capabilities through simulation, SIL/HIL, and real-world flight test
Ability to communicate technical results clearly to engineering, autonomy, and program stakeholders
Ability to operate independently, take ownership of full algorithm lifecycle, and iterate rapidly against field feedback
Comfortable working in an iterative R&D environment with rapidly changing requirements and priorities
U.S. Person status required to support ITAR-controlled programs

Nice To Haves

Master's or PhD in Computer Vision, Machine Learning, Robotics, or related field
Prior experience developing perception or autonomy for UAS, robotics, autonomous vehicles, or defense platforms
Experience with on-vehicle perception on dynamic platforms (high angular rates, rapid scene changes, motion blur)
Experience with visual-inertial odometry (VIO), visual SLAM, and GPS-denied navigation
Experience with single- and multi-object tracking (MOT), re-identification, and tracking through occlusion or sensor handoff
Experience developing target lock, terminal guidance, or last-mile autonomy systems
Experience with sensor fusion across camera, IMU, GNSS, radar, LiDAR, or thermal modalities
Experience with thermal/IR imagery, low-light imaging, event-based vision, or stereo/RGB-D systems
Experience with GPU/CUDA programming for accelerated computer vision processing
Experience training and deploying foundation models, vision-language models, or self-supervised learning approaches
Experience operating in GPS-denied, contested, or EW-affected environments
Experience with synthetic data generation, sim-to-real transfer, or domain randomization (Unreal/Unity/Isaac Sim, Gazebo)
Familiarity with NVIDIA Jetson Orin, Hailo, Ambarella, Qualcomm RB-series, or similar edge platforms
Familiarity with DoD test ranges, COA operations, or military customer environments
Active or prior security clearance
Experience working in fast-paced defense or dual-use technology environments
Publications, open-source contributions, or competition results (KITTI, COCO, nuScenes, etc.) demonstrating CV expertise

Responsibilities

Design, train, evaluate, and deploy computer vision models for detection, classification, segmentation, tracking, and re-identification of ground and aerial targets
Develop and harden visual-inertial odometry (VIO), visual SLAM, and GPS-denied navigation capabilities for tactical FPV operations
Build and maintain target lock, terminal guidance, and last-mile autonomy perception pipelines aligned with operational mission requirements
Write custom computer vision algorithms from scratch where library implementations don't meet platform constraints (real-time, embedded, adversarial conditions)
Own camera and sensor integration: calibration workflows, intrinsic/extrinsic estimation, multi-sensor synchronization, and image pipeline development across RGB, thermal/IR, and stereo modalities
Optimize models for real-time inference on embedded edge compute platforms (NVIDIA Jetson, Hailo, Ambarella, Qualcomm, or similar) using TensorRT, ONNX, quantization, pruning, and distillation techniques
Write production-quality C++ and Python inference code integrated with onboard flight software and autonomy stacks; apply GPU/CUDA programming for accelerated processing where required
Curate, label, and manage flight-collected datasets; build data pipelines for ingestion, augmentation, versioning, and retraining
Establish evaluation frameworks and validate perception capabilities across the full test stack: unit tests, simulation, software-in-the-loop, hardware-in-the-loop, and live flight testing
Collaborate with autonomy, flight software, and hardware teams to define camera, sensor, and compute requirements; make pragmatic engineering tradeoffs under SWaP constraints
Support flight-test campaigns by analyzing onboard video, telemetry, and inference logs to diagnose model failures and prioritize improvements
Investigate and integrate emerging techniques in foundation models, multimodal perception, sensor fusion, and on-device learning where they advance platform capability
Harden perception against degraded visual conditions (low light, dust, motion blur, occlusion) and against GPS-denied, EW-contested, and visually adversarial environments
Document algorithms, models, and pipelines to support team velocity, reproducibility, and customer deliverables
Operate effectively in outdoor field environments, including supporting data collection events and live flight testing as required