Computer Vision Engineer

GenLogs•Washington, DC

About The Position

The Data Science team at GenLogs transforms raw observational data from our sensor network into high-value intelligence used by law-enforcement agencies, regulators, ports, and private-sector freight operators. We build models, analytics, and measurement frameworks that enable vehicle detection, entity resolution, behavioral insights, fraud and theft indicators, compliance signals, and network-wide operational performance metrics. Our work sits at the center of our platform, shaping how billions of roadside observations become actionable information. We partner closely with Engineering and Product to deploy algorithms at scale and with Go-to-Market teams to define customer-facing analyses that drive real operational outcomes. The team blends statistical rigor, ML capability, and domain expertise to create a new standard for freight intelligence in the United States. You will own computer-vision problems end to end from the photons entering a roadside camera to the structured vehicle intelligence delivered by our platform. This is not a role where you train a model, publish an evaluation, and hand it to another team to productionize. You will be responsible for understanding the entire system: camera placement and configuration, image quality, training data, model architecture, edge inference, production deployment, monitoring, and downstream outcomes. Our operating environment is unforgiving. Trucks move at highway speeds through darkness, glare, rain, snow, occlusion, extreme perspectives, and inconsistent connectivity. Models must run reliably on constrained edge hardware across a geographically distributed sensor network. Improvements that look promising offline must survive actual roadside conditions and measurably improve the intelligence our customers receive. You will have the autonomy to attack these problems wherever the evidence leads. One day that may mean designing a better OCR or detection model. The next may mean profiling a TensorRT pipeline, diagnosing video compression artifacts, redesigning an evaluation dataset, or working with field teams to change exposure, shutter speed, lighting, or camera positioning. We’re looking for someone who doesn’t stop at “the model works.” You own the outcome.

Requirements

Proficient in Python, SQL, OpenCV, and PyTorch programming
Proficient in fundamental computer vision techniques including planar homography, keypoint detection, perspective projection, epipolar geometry, object tracking, camera calibration (intrinsic/extrinsic parameters), and re-identification
Knowledgeable in GPU computing and inference optimization tools including CUDA, ONNX, OpenVINO, and TensorRT
Knowledgeable in object detection, segmentation, optical character recognition, signal processing, linear algebra, and Vision Transformer models
Experience or interest in geospatial data analysis (e.g., GIS, GeoPandas)
Experience with version control systems like Git
Experience with linux systems
Minimum 4 years of professional experience in computer vision, data science, or research
Hands-on experience with applied computer vision techniques, especially on edge hardware
Strong problem-solving skills and attention to detail
Excellent written and verbal communication skills
Ability to communicate technical concepts to both technical and non-technical stakeholders

Nice To Haves

Own outcomes, not artifacts: your responsibility does not end with a trained model, a notebook, or a pull request.
Follow problems across boundaries: you will move between research, data, production software, hardware, and field operations when the problem requires it.
Are comfortable being the expert in the room: you can form a point of view, defend it with evidence, and make consequential technical decisions.
Operate well without a playbook: many of our hardest problems do not have established benchmarks or obvious solutions.
Prefer reality over elegant abstractions: you care whether the system works on a dark highway in the rain, not merely whether it performs well on a curated dataset.
Build durable systems: you consider monitoring, failure recovery, maintainability, and operational performance part of the work.
Have urgency without sacrificing rigor: you iterate quickly, measure honestly, and know when the evidence is strong enough to ship.
Finish the job: when something fails, you stay with it until the underlying problem is understood and resolved.

Responsibilities

Own computer-vision capabilities from initial problem definition through field deployment and sustained production performance.
Build systems for vehicle detection, tracking, OCR, attribute extraction, embedding generation, and re-identification.
Establish evaluation frameworks that connect model performance to successful vehicle identification and customer outcomes, not merely offline accuracy.
Diagnose failures across the entire imaging pipeline, including camera placement, optics, exposure, lighting, encoding, training data, model behavior, edge compute, and downstream processing.
Develop models and write the production software required to run them reliably at scale.
Optimize multi-model pipelines for latency, throughput, memory utilization, and accuracy on constrained CPU and GPU edge hardware.
Design experiments, analyze failure modes, and make clear decisions about what to improve next.
Build monitoring and feedback loops that reveal model degradation, environmental shifts, hardware issues, and new failure modes.
Work directly with field operations to test camera configurations and validate improvements under real roadside conditions.
Partner with platform and data engineers while remaining accountable for getting your work into production.
Improve annotation strategy, training datasets, and hard-example selection based on observed production failures.
Evaluate new research, determine what is genuinely useful, and translate promising ideas into dependable systems.
Raise the technical standard for computer vision across GenLogs through strong engineering, documentation, and mentorship.

Benefits

Employer-covered comprehensive medical, dental, and vision plans
Employer contribution towards premiums of optional higher-end plans
Unlimited PTO
Sick leave
Company holidays (GenLogs observes all US Government holidays)
Flexible leave for caregiving and medical needs
Paid parental leave
Budget availability for approved professional development courses, certifications, and training
100% travel reimbursement for all approved company travel and spending
401(k) plan (US based employees)