Machine Learning Engineer (Mid-Level) – Computer Vision Remote Sensing

Role Overview

You will build ML and computer vision components for a plantation health monitoring platform that fuses satellite, aerial, drone, and in-field imagery. The scope of the work includes geospatial data preparation, model development for tree and disease detection, and forecasting models for yield and agronomy recommendations. You will independently own well-scoped modules end-to-end and contribute to design discussions around data, modeling, and productionization.

Key Projects

Plantation Health Monitoring Platform
A multi-modal monitoring system that aggregates satellite imagery, aerial surveys, drone captures, and field camera images to assess plantation conditions at scale. The platform supports disease detection, tree detection and counting, yield forecasting, and recommendations for fertilization and irrigation levels. Engineering challenges include geospatial data alignment across sources, dataset quality control, model robustness across seasons and sensors, and deploying repeatable pipelines for continuous monitoring.

Responsibilities

• Build and maintain data pipelines to ingest, normalize, and align multi-source imagery (satellite, aerial, drone, field cameras) into ML-ready datasets.

• Implement CV models for tree detection and health/disease signal detection, including training, evaluation, and error analysis across regions and seasons.

• Develop forecasting features and models for yield prediction using aggregated signals from imagery and derived indices (like NDVI, EVI, or RECI).

• Design practical agronomy-related predictors (e.g., fertilizer and irrigation need estimation) by translating domain outputs into measurable ML targets and validation protocols.

• Create dataset QA workflows (labeling guidelines, sampling strategies, drift checks) to keep training data consistent as new farms and sensors are added.

• Optimize inference pipelines for batch and near-real-time processing where needed (throughput, cost, and reliability).

• Contribute to interface contracts between ML components and the broader platform (data schemas, model outputs, versioning).

• Participate in code reviews, write tests for data/model logic, and document model assumptions, limitations, and monitoring signals.

Required Qualifications

• 3-5 years of hands-on experience building ML systems, including shipping at least two model or ML pipeline into a production or production-like environment.

• Strong Python skills and experience with at least one deep learning framework (PyTorch preferred, TensorFlow acceptable).

• Practical CV experience: detection/segmentation fundamentals, augmentation, evaluation metrics (mAP, IoU/F1), and systematic error analysis.

• Core Frameworks – PyTorch (preferred) or TensorFlow; experience with custom training loops, mixed precision (FP16), and experiment tracking.

• Detection & Segmentation – hands-on experience with modern architectures for detection/segmentation such as YOLO, RT-DETR, Mask R-CNN, and U-Net/DeepLab-style segmentation for canopy/health mapping.
• Deployment & Optimization – practical experience with TensorRT, OpenVINO, or ONNX Runtime for inference acceleration.
• Multi-source Data Fusion – developing pipelines to resolve inconsistencies in resolution, viewpoint, lighting, seasonality, and sensor noise across satellite/aerial/drone/ground imagery.
• Geospatial & Remote Sensing – raster processing with GDAL/rasterio; vector ops with GeoPandas/Shapely; working with GeoTIFF/COG, coordinate reference systems (CRS), tiling/patching large rasters; basic spatial joins and alignment.

• Production ML Basics – containerized development (Docker), Linux workflows, reproducible pipelines, dataset versioning concepts, and performance-aware preprocessing.

• CS/EE/Math degree or equivalent practical experience.

Nice to Have

• Foundation vision models – experience using or fine-tuning SAM, DINOv2, or Grounding DINO for annotation acceleration, zero-shot/few-shot bootstrapping, or improved segmentation.

• Time-series forecasting – feature engineering and evaluation for yield forecasting; experience with models like XGBoost/LightGBM or deep forecasting (e.g., Temporal Fusion Transformer / N-BEATS).

• Experience with STAC catalogs or cloud-native geospatial patterns for large-scale imagery access.

• Experience designing labeling workflows for geospatial imagery (active learning, weak supervision, consensus labeling).

Working conditions:

• We provide an inspiring working environment where our employees feel rewarded and engaged.
• We expect a lot from our employees and are ready to give a lot in return. You’ll be faced with challenging, varied, non-standard projects and tasks. But at the end of the day, you’ll be proud of what you’ve done.
• We strongly encourage the growth and development of our team. It is in your best interest to learn new languages and technologies and to implement them into existing and new projects. It won’t be unattended, and we will definitely reward you.
• We pay a lot of attention to the health of our employees, so we offer comprehensive health insurance that also covers dental services. So drink tea with ginger and lemon, we have it year-round in the office kitchen.
• Softarex Technologies treats each employee individually. Our HR team helps newcomers at every stage of adaptation in the company. No less attention is paid to employees who feel at home here (they literally have their own slippers).

Of course, that’s not all. Check out the full benefits package here and let’s get started!