Computer Vision Engineer

07 Oct Computer Vision Engineer

About Black Swift Technologies

Founded in 2011 and based in Boulder, Colorado, Black Swift Technologies is a premier engineering firm and pioneer in specialized unmanned aircraft systems (UAS) built to operate in the world’s most demanding and hazardous atmospheric conditions. From intercepting tornadic supercell storms in the VORTEX2 field campaign and sampling active volcanoes to measuring soil trafficability and wildland fire behavior, Black Swift delivers highly durable, American-made dual-use aerial platforms. All Black Swift systems are anchored by the proprietary, custom-designed SwiftCore™ Flight Management System (FMS), which encompasses high-performance autopilots, portable ground stations (SwiftStation™), and robust control electronics designed from the ground up to ensure strict NDAA and ITAR compliance. By controlling the entire design, hardware, and avionics supply chain, Black Swift guarantees the quality and operational resilience of its aircraft, such as the air-deployed SØ dropsonde-mimicking glider and the long-endurance E2 soil moisture mapping drone. As a key operating group within the KrateoSky autonomous intelligence network, Black Swift Technologies leverages common platforms, unified perception systems, and AI-native software architectures to deliver critical aerial capabilities across defense, public safety, and scientific research missions.

Position Summary

You will build onboard computer-vision capabilities across Black Swift aircraft, payloads, and autonomy workflows. Your job is to turn raw imagery into reliable, mission-useful information on real systems under real constraints. That means detection, tracking, scene understanding, and image or video analysis are only the beginning; the actual job is getting those capabilities to run consistently on constrained edge hardware, inside an operational software stack, with performance that survives field conditions and engineering scrutiny.

High-agency IC role: This is a high-agency individual contributor role. You will be hands-on from day one. We are looking for an engineer who writes model code, builds and profiles inference pipelines, debugs sensor and integration issues, and pushes systems all the way from promising prototype to fieldable runtime. You do not stop at offline metrics. You care about latency, memory footprint, thermal behavior, fault modes, integration quality, and whether the system actually performs on aircraft rather than in a notebook.

You are AI-native in how you work—AI assistants are your default environment. You use LLMs and coding copilots to generate boilerplate, accelerate experimentation, write tests, build profiling harnesses, compare architecture options, and compress iteration loops. You see AI as leverage for a small, elite engineering team, not as a gimmick. You are still expected to think clearly, verify aggressively, and own the output.

Embody our values and behaviors—deliver precision and quality, challenge how things are done, break problems into fundamentals, automate relentlessly, hold yourself accountable, and deliver together no matter the obstacles. You must exhibit Above the Line behaviors: operating as a Coach, Creator, and Challenger rather than a Victim, Hero, or Villain. You apply automation-first engineering: your default instinct is to script repetitive work, systematize evaluation, and eliminate manual friction rather than tolerate it. You operate as an egoless collaborator with strong documentation discipline, and you maintain a PDCA (Plan-Do-Check-Act) orientation—willing to test, learn, correct, and iterate quickly.

Key Responsibilities

• Onboard Perception Development: Design, train, evaluate, and refine computer-vision models for onboard target detection, tracking, classification, and scene interpretation across the Black Swift aircraft and payload stack.
• Edge Runtime Optimization: Build and maintain robust inference pipelines that run on constrained onboard compute, optimizing for latency, throughput, memory usage, thermal limits, and reliability on hardware that does not forgive waste.
• Model-to-System Integration: Work directly with autonomy, embedded, software, and payload engineers to integrate perception capabilities into the broader flight and mission software environment rather than treating computer vision as an isolated research surface.
• Hardware-Aware Engineering: Tune models and pipelines for edge accelerators and constrained compute environments such as Jetson-class devices, CPUs, FPGAs, TPUs, NPUs, or equivalent low-SWaP platforms, making principled trade-offs between accuracy, power, memory footprint, and execution speed.
• Validation Against Real Mission Criteria: Develop repeatable evaluation workflows that measure accuracy, false positives, robustness, and timing performance against actual operational requirements rather than generic benchmark vanity metrics.
• Flight and Field Debugging: Support bench, flight-test, and post-flight analysis activities to identify failure modes, trace perception defects through the software and sensor stack, and convert findings into hard engineering fixes.
• Dataset and Experiment Discipline: Help structure labeling, dataset quality, experiment tracking, regression baselines, and test harnesses so that model improvements are measurable, reproducible, and durable over time.
• AI-Native Execution: Use AI tools aggressively but responsibly to accelerate code generation, evaluation harnesses, optimization passes, and debugging while maintaining a high bar for correctness and engineering rigor.
• Reusable Architecture Building: Contribute patterns, tooling, and documentation that make perception capabilities easier to scale, maintain, and reuse across future platforms rather than creating one-off brittle implementations.
• Cross-Functional Delivery: Partner closely with technical leadership and adjacent engineering teams to resolve interface issues, clarify trade-offs, and keep perception work aligned with delivery needs, not just technical curiosity.

Required Qualifications

• Engineer II Level Delivery: Typically 3–5 years of relevant experience, or equivalent demonstrated capability, in computer vision, machine learning, robotics, autonomy, or a closely related domain.
• Proven Track Record: You have personally built, optimized, and shipped computer-vision systems that ran on physical hardware in production, field, robotic, or UAS-adjacent environments. We are not looking for someone whose experience ends at experimentation.
• Technical Depth: Strong fluency in Python and C++, with practical command of modern CV/ML tooling such as PyTorch, ONNX Runtime, TensorRT, OpenCV, or comparable frameworks used for real deployment rather than toy examples.
• Embedded and Edge Mastery: Direct experience deploying models to embedded or edge hardware and making hard trade-offs around compute, memory, thermal behavior, and runtime performance.
• Linux / Target-System Fluency: Comfortable working in terminal-driven Linux/POSIX environments for deployment, profiling, integration, and debugging on target and edge systems such as Jetson-class compute.
• Systems and Troubleshooting Depth: You can trace a failure across sensor input, preprocessing, model inference, post-processing, integration logic, and downstream software behavior and drive the issue to resolution.
• UAS Domain Familiarity: Practical understanding of how perception fits into aircraft, payload, and autonomy environments, including the realities of onboard compute, flight software integration, and engineering validation on real systems.
• AI-Native Proficiency: Cursor, GitHub Copilot, and similar assistants are part of your normal workflow; you are highly productive using AI to accelerate code, tests, scripts, profiling, and debugging.
• Clearance Eligibility: Due to the nature of this role and the work performed, candidates must be eligible to obtain and maintain a U.S. Department of Defense (DoD / Department of War) Secret security clearance.

Desired Qualifications

• Experience with aerial, remote-sensing, or geospatial imagery.
• Experience with model optimization for accelerators or other low-SWaP compute environments.
• Familiarity with visual navigation, visual inertial odometry, or adjacent perception-driven autonomy workflows.
• Familiarity with defense-adjacent, regulated, or high-assurance engineering environments such as Blue UAS, NDAA-conscious sourcing, or related compliance boundaries.

Logistics

• Department: R&D Engineering / Autonomy
• Location: Boulder, CO
• Reports To: Maciej Stachura, Ph.D. (CTO)
• Compensation: $120,000 – $145,000 / year (targeted base; total package includes comprehensive benefits)