Machine Learning Engineer (Computer Vision)

This role focuses on computer vision for extracting insights from imagery at scale, alongside, * Build computer vision models for segmentation, detection, and classification

• Determine appropriate modeling approaches (deep learning vs. classical image

processing)

• Develop NLP and LLM agent solutions for structured data extraction

• Create labeling guidelines and construct ground-truth datasets

• Establish evaluation suites, test sets, and monitoring for models and agents

• Conduct quick feasibility studies and research paper implementations

• Collaborate with engineers to deploy scalable production models

Machine Learning Foundations

• Strong ML fundamentals: linear algebra, backprop, optimization, PyTorch, and NumPy

• Understanding model failure modes (overfitting, leakage, class imbalance, distribution

shift)

Computer Vision Foundations

• Image geometry, projection, homographies, camera models, and classical image

processing

• Deep learning for vision: CNNs, segmentation (e.g., U-Net), object detectors, and ViTs

• Evaluation metrics (IoU, mAP, precision/recall) and custom dataset training experience

NLP & LLM Agents

• Understanding of transformers, embeddings, and tokenization

• Experience with LLM agents: tool calling, retrieval, structured output, and context

management

Evaluations & Tracking

• Experience designing slice-based evals, agent testing benchmarks, and LLM-as-judge

frameworks

• Proficiency tracking experiments and model versions using tools like W&B or MLflow

Work Style & Approach

• Strong problem decomposition and clear technical communication

• Effective leverage of AI coding tools with critical verification

Preferred / Bonus Qualifications

• Backend engineering: you’ve built or run APIs and data services (we use Python,

FastAPI and Postgres)

• Distributed systems: you’ve spread work across many machines or GPUs and dealt with

what goes wrong, like job queues, retries and partial failures

• Papers in reputed international conferences or journals, e.g. CVPR, ICCV, ECCV,

WACV, NeurIPS, ICLR, ACL, EMNLP, and remote sensing venues like IGARSS, ISPRS

or IEEE TGRS

• Observability: instrumenting services and agents with OpenTelemetry, including the

GenAI semantic conventions for model and tool calls, and monitoring models for drift in

production

• Recent vision foundation models: SAM 3, DINOv3, open-vocabulary detectors, and

VLMs used for zero-shot classification or pre-labeling

• Remote sensing or geospatial work: multispectral imagery, NDVI, GeoTIFFs, coordinate

systems, GDAL/rasterio, PostGIS, and geospatial foundation models like Prithvi, Clay or

AlphaEarth

• GPU performance and inference cost: profiling, batching, mixed precision, quantization,

distillation, TensorRT or ONNX Runtime

• Docker and Kubernetes

• Competitive salary determined based on experience and interview performance.

• Hybrid role with remote work flexibility, with one or two potential in-person weekly

meetings.

• Paid vacation days, sick days, and holidays for full-time employees.

• Knowledge-sharing sessions on the latest in tech, engineering, and design.

Oak Network builds decentralized infrastructure for crowdfunding and commerce, enabling

creators to run campaigns and shops with seamless blockchain-backed trust and payments.

We are a fully remote global team where members take ownership and collaborate across time