Founding Computer Vision and ML Engineer / Tech Lead — Earth Observation

About Satlantis

Satlantis builds high-performance satellite systems and the data-processing pipelines that turn what they see into decisions. Our Earth-observation payloads and analytics serve customers in environmental and infrastructure monitoring, maritime awareness, and defense, where image quality and model accuracy have direct consequences. Our Gainesville team owns data processing, deployed analytics, and intelligence products, plus the tooling that keeps throughput high. More at www.satlantis.com.

The Role

You will build and run the CV/ML function at Satlantis US: the technical direction for imagery understanding and applied ML across our Earth-observation products, and the team that delivers it. You report to a CTO with deep ML expertise, so your decisions get a sparring partner rather than a translation layer.

Computer vision is the core of the work, along with the dataset and evaluation discipline that makes it trustworthy. Adjacent applied ML sits with you too, so the role suits deep vision expertise paired with broad ML judgment. You will work most closely with the image processing and mission teams at our headquarters in Spain: peers, not a reporting line. You set the U.S. roadmap and align on shared standards, tooling, and datasets so the group builds once rather than twice.

This is a hands-on role: expect around half your time prototyping hard problems, reviewing architectures, and writing code where it matters most. The other half is what only the lead can do: setting a roadmap that survives contact with mission requirements, hiring and growing engineers, making build-versus-buy calls, and representing CV/ML to product, mission, and customers. We are not looking for someone who stops being an engineer.

What You'll OwnTeam and people leadership

• Build the team. Own hiring for CV and applied ML: defining roles, calibrating the interview loop, closing candidates, and onboarding them onto real work quickly.
• Grow engineers. Set expectations, give direct feedback, run one-on-ones and performance conversations, coach through design and code review, build career paths that keep strong ICs technical, and delegate ownership so engineers decide well without you.
• Set the culture. Establish standards for reproducibility, evaluation rigor, documentation, and peer review, then model them.

Technical direction

• Set the CV/ML roadmap. Own a 12–18 month plan: imagery understanding (segmentation, detection, classification, image matching, semantic retrieval, change detection, tracking, anomaly detection), the adjacent applied ML the business needs, and the automation that delivers new imagery products faster, sequenced against mission requirements and real capacity.
• Turn requirements into tractable problems. Convert customer and mission needs into well-scoped problems with explicit metrics (precision/recall, mAP, IoU, F1, latency, throughput, memory footprint), acceptance criteria, and validation plans.
• Own dataset strategy. Annotation protocols, tiling and sampling, class balance, hard-negative mining, augmentation policy, domain-shift analysis, label-quality audits, and dataset versioning and documentation.
• Choose the architecture bets and the trade-offs. Guide use of CNNs, vision transformers, encoder-decoder designs, self-supervised learning, multi-modal fusion, and foundation-model adaptation, under real constraints like ground sample distance, viewing geometry, atmospheric effects, and multi-temporal data. Decide what to build, buy, or leave alone, and document the reasoning.

Delivery and operations

• Ship into production. Partner with software and platform engineering on model packaging, inference optimization, deployment, monitoring, drift detection, versioning, and rollback. Own the accuracy, latency, reliability, and cost of deployed systems, not just the offline benchmark.
• Get the most from the compute we have. Model training runs on HiPerGator, UF's AI supercomputer, under a fixed annual allocation, so the lever is utilization rather than spend: scheduling and queue discipline, data-loading efficiency, and more experiments from the same capacity.
• Make the pipeline faster. Raise data-processing throughput and shorten the path from capture to delivered product: automated image quality checks, reusable evaluation and release tooling, and fewer manual steps.

Working across the group

• Be the U.S. technical counterpart. Build working relationships with the image processing, mission, and engineering teams across the group: share results, review each other's approaches, and know what exists elsewhere before building it here. Converge on evaluation methodology, dataset conventions, model packaging, and tooling where it pays, and say plainly when local needs differ.
• Operate without shared hours. Sustain momentum across a six-hour time difference with clear written communication, durable design documents, and asynchronous decisions. Make U.S. work reusable in the other direction.

What Success Looks Like

• First 30 days. You can say where the real bottlenecks are in our imagery, pipelines, and model performance, and you have met the stakeholders who depend on them.
• First 90 days. A prioritized roadmap is agreed with U.S. leadership and shared with the other teams involved across the group, evaluation practices are consistent, and at least one meaningful improvement is in production.
• First year. The team has grown deliberately, ships predictably, and owns workstreams independently. CV/ML is a differentiator in customer conversations, U.S. and group efforts reinforce rather than overlap, and you have a plan for the next 12 months.

What We're Looking ForRequired

• Degree in computer science, computer or electrical engineering, remote sensing, robotics, or a related field, or equivalent practical experience.
• 5+ years in computer vision, machine learning, or applied AI, with vision models delivered into production or operational use.
• 2+ years leading engineers as a manager or technical lead, accountable for what the team delivered, including hiring, onboarding, developing talent, and giving difficult feedback well.
• Deep command of CV fundamentals: image representations, feature extraction, geometric reasoning, dense prediction, detection, segmentation, and evaluation methodology.
• Strong Python and hands-on depth in PyTorch (preferred) or TensorFlow, with the ability to still write clean, tested, maintainable code.
• Fluency in training and inference optimization (data-loading efficiency, batching, mixed precision, model compression), applied to large-scale imagery datasets and reproducible pipelines.
• Able to explain a technical trade-off to a mission lead, a customer, and an executive without losing any of them.

Preferred

• Geospatial and remote sensing: GDAL, Rasterio, projections and CRS handling, tiling strategies, GeoTIFF/COG/NetCDF, STAC/PgSTAC, image-quality assessment, multi-spectral and panchromatic imagery, super-resolution, image fusion, orthorectification-aware workflows, change detection.
• Spatiotemporal modeling: time-series imagery, temporal fusion, motion and change analysis, event detection, tracking across repeated captures.
• MLOps and data governance: model serving, monitoring, experiment tracking (W&B, MLflow, CometML), orchestration (Airflow, Argo, ZenML), lineage, and dataset documentation.
• Cloud and HPC at scale: AWS/GCP/Azure, on-prem clusters, SLURM-managed GPU fleets, Kubernetes, containers, distributed training.
• Foundation models for vision or Earth observation: fine-tuning, embedding extraction, retrieval, transfer learning, promptable models, multimodal representation learning.
• Performance-oriented deployment: C++, OpenCV, ONNX, TensorRT, Triton, CUDA optimization, edge or real-time inference.
• Also useful: owning a budget, vendor relationships, or academic partnerships; multinational experience coordinating with peer teams at another site; working Spanish, though all technical work is in English.

Location, Authorization, and Compensation

Full-time and on-site in Gainesville, Florida. Leading this team means being in the room with the engineers and product colleagues you work alongside, so we cannot offer remote work.

MUST BE A U.S. CITIZEN. The selected candidate must be eligible to obtain and maintain a Secret security clearance. This position does not provide employment visa sponsorship; candidates must possess permanent, unrestricted legal authorization to work in the U.S.

Salary is competitive and commensurate with experience, plus a performance-based bonus. Benefits include medical, dental, and vision coverage and paid time off.

Why Satlantis

• Own a function, not a ticket queue: technical direction for CV and applied ML here is yours to set.
• Imagery and problems few teams get access to, with tight feedback loops between data, model, and customer.
• Small enough that your decisions land quickly; established enough that the satellites are already flying.
• Colleagues genuinely good at hard things: software engineering, cloud infrastructure, high-performance computing, geospatial data, and ML in one team, plus experienced image processing, mission, and engineering teams group-wide.
• Gainesville: a university town with low cost of living and quick access to both coasts (visitgainesville.com).

How to Apply

Send your resume and a short note to [email protected]. Tell us about a vision system you led into production: what you decided, what you got wrong, and what you would do differently.

Satlantis is an equal opportunity employer. We consider all qualified applicants without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, protected veteran status, or any other legally protected characteristic.