Computer Vision Engineer (Visual Place Recognition)

About us

Farsight Vision converts flight footage into digital 2D and 3D twins for real-time intelligence in GNSS-denied environments, making analytics and situational awareness convenient and accessible while saving time and effort. We create multi-layered digital twins of terrain with dynamic tracking and object/landscape monitoring and predicting.

As a Computer Vision Engineer focused on Visual Place Recognition (VPR), you will own the models and pipelines that answer a critical question in the field: where are we? You will research, train, and ship VPR systems that work under viewpoint change, illumination shift, seasonal variation, and degraded imagery typical of real operations - not only clean academic benchmarks.

If you thrive in startup environments, are proactive, and enjoy turning research into reliable production systems, this role is a chance to shape core localization capability for a defense-tech platform used in live operations.

Responsibilities

• Own Visual Place Recognition end-to-end: problem framing, dataset curation, training, evaluation, and deployment into the FSV Platform.

• Design, train, and iterate on VPR models and retrieval pipelines - global descriptors, local feature matching, and hybrid approaches as the task requires.

• Work with and extend modern VPR / visual localization methods (for example NetVLAD, MixVPR, CosPlace, EigenPlaces, AnyLoc, DINOv2-based descriptors, SuperPoint / SuperGlue / LightGlue-style matching) and know when classical or learning-based methods fit better.

• Build robust training pipelines: data prep, augmentations for domain shift, hard-negative mining, loss design, hyperparameter search, and reproducible experiment tracking.

• Evaluate models beyond top-line recall - failure modes under night, blur, compression, seasonal change, and map-query domain gap; define metrics that matter for field use.

• Package and serve models for production inference (GPU and constrained / edge settings when needed), working with backend and MLOps on versioning, monitoring, and redeployment.

• Collaborate with backend, geospatial, and field teams to integrate VPR into mapping, localization, and digital-twin workflows.

• Use AI coding tools daily, and review, test, and stand behind every line and every model that ships.

• Run small PoCs on new architectures or training recipes, measure them honestly, and bring the team a reasoned recommendation.

Requirements

• 3+ years in computer vision / deep learning for real products or research transferred to production.

• Strong Python and hands-on experience with PyTorch (or equivalent) for training and debugging vision models.

• Proven experience with Visual Place Recognition, visual localization, image retrieval, or closely related geo-localization / SLAM-adjacent perception.

• Ability to train and fine-tune VPR / retrieval models from scratch or from public checkpoints: datasets, losses, mining strategies, and evaluation protocols (e.g. recall@N, precision-recall under realistic splits).

• Working knowledge of modern VPR and matching literature and the judgment to pick methods for viewpoint, appearance, and domain shift - not only recreate paper numbers.

• Experience with large-scale feature indexes / ANN search (FAISS or similar) and practical retrieval system design.

• Solid understanding of CNN and transformer backbones, representation learning, and metric learning.

• Comfort debugging models and data issues that show up in the field (noisy labels, distribution shift, rare failure cases).

• Experience in a fast-paced environment with high ownership and ambiguous requirements.

Nice to have

• Published work or strong open-source contributions in VPR, visual localization, or image retrieval.

• Multi-view geometry, SfM / SLAM, or photogrammetry experience (COLMAP, OpenMVG, etc.).

• Geospatial stack: georeferencing, map alignment, GDAL, orthophotos, point clouds.

• Model optimization for deployment: ONNX, TensorRT, quantization, distillation.

• Video / sequential place recognition, temporal consistency, or map maintenance over time.

• Experience with aerial / drone imagery and GNSS-denied or contested environments.

• CUDA and GPU training at scale.

• Defense tech or dual-use background.

Security Eligibility

Our platform is actively used by the Armed Forces in live operations. As an Estonian-Ukrainian defense technology company handling sensitive operational data, we have a responsibility - legal, ethical, and practical - to ensure the integrity of everyone on our team.

All candidates must be able to confirm the following:

• Neither you nor your close relatives (parents, spouse, siblings, children) reside in Russia, Belarus, or temporarily occupied Ukrainian territories

• You have no regular personal, financial, business, or professional contact with individuals in those territories

• You consent to a deeper security verification if the process advances to later stages

If any of the above criteria are not met, please do not proceed with your application.

Why us

• With us, you will become part of a team of professionals who strive to contribute to the development of advanced technologies.

• We value teamwork and create an atmosphere of mutual support where everyone can unleash their potential.

• Flexible work schedule.

• Paid vacation and sick days.

Hiring process

• Technical Interview (CV / VPR deep dive + practical discussion of training and evaluation).

• PM & C-level management Interview.