Computer Vision Developer

Skyshi Digital Indonesia; Jakarta Pusat; Hybrid; IDR 200,880,000 - 323,640,000; Full time

Job summary

Skyshi Digital Indonesia seeks a Computer Vision Engineer to assess the feasibility of offline, on-device image recognition and to build and deploy YOLO-based detectors end-to-end.

You will also evaluate cloud CV services (AWS Rekognition) and optimize models for edge deployment, including quantization and format conversions, while coordinating with product and engineering teams.

Qualifications

• Hands-on CV experience with YOLO or similar detection architectures.
• Experience taking offline models from research to production/edge deployment.
• Familiarity with cloud CV services (AWS Rekognition or equivalent).
• Understanding of MLOps basics: model versioning, monitoring, retraining pipelines.
• Proficient in Python and CV tools (OpenCV, PyTorch/TensorFlow).
• Able to assess feasibility, effort, and risk before projects start.

Responsibilities

• Assess offline image recognition feasibility: architecture, dataset size, hardware, latency, storage.
• Design, train, and optimize YOLO-based detection models for offline edge deployment.
• Prepare data, annotate, augment, and evaluate models (precision, recall, mAP).
• Optimize models for deployment (quantization, pruning, ONNX/TensorRT/TFLite).
• Handle end-to-end deployment to production/edge devices with monitoring.
• Explore cloud CV solutions as alternatives or complements to offline models.
• Provide trade-off recommendations between offline vs cloud-based approaches.
• Collaborate with product/engineering and document experiments and results.

Skills

Problem solving

Communication

Cross-functional collaboration

Feasibility assessment

Education

Bachelor's degree in Computer Science, Engineering, Mathematics, or related field

Tools

Python

PyTorch/TensorFlow

OpenCV

ONNX Runtime

Docker

Job description

Overview

We are looking for a Computer Vision Engineer who can assess the technical feasibility of offline (on-device / edge) image recognition solutions, build and deploy YOLO-based detection models end-to-end, and also work with cloud-based computer vision services such as AWS Rekognition. Prior hands‑on experience taking an offline model from research to production deployment is highly valued.

Key Responsibilities

• Conduct technical feasibility assessments for offline image recognition needs: architecture selection, dataset size, compute/hardware requirements (edge device, GPU/CPU), latency, and storage constraints before development begins.
• Design, train, and optimize image recognition/object detection models based on YOLO that can run fully offline (on-device/edge, no dependency on internet/cloud connectivity).
• Perform data preparation, annotation, augmentation, and model evaluation (precision, recall, mAP, etc.).
• Optimize models for deployment (quantization, pruning, conversion to lightweight formats such as ONNX / TensorRT / TFLite) so they run efficiently on target devices.
• Handle end‑to‑end deployment of computer vision models to production/edge devices, including post‑deployment monitoring and maintenance.
• Explore and implement cloud‑based computer vision/image recognition solutions (e.g., AWS Rekognition, Azure Computer Vision, Google Vision AI) as alternatives or complements to offline solutions.
• Provide trade‑off recommendations between offline (custom model) and cloud‑based (managed service) solutions based on business needs, cost, data privacy, and infrastructure conditions.
• Collaborate with product/engineering teams to integrate CV solutions into broader systems.
• Document research processes, experiments, and evaluation results in a structured way.

Qualifications

• Hands‑on experience building computer vision/image recognition models using YOLO (or similar detection architectures).
• Proven, real experience building offline models — from research and training through to production/edge deployment (not just notebook‑level experiments).
• Familiarity with cloud‑based CV services such as AWS Rekognition (or equivalent), able to compare use cases against custom/offline solutions.
• Understanding of basic MLOps: model versioning, performance monitoring, retraining pipelines.
• Familiar with tools/frameworks: Python, PyTorch/TensorFlow, OpenCV, ONNX Runtime, Docker.
• Able to independently perform technical assessments (feasibility, effort, and risk) before project execution.

Nice to have:

• Experience with edge devices (Jetson, Raspberry Pi, etc.) or mobile deployment (TFLite/CoreML).

Requirements

• Bachelor's degree in Computer Science, Engineering, Mathematics, or a related field.
• Minimum 2-5 years of hands‑on experience in computer vision/machine learning.
• Demonstrated portfolio or GitHub of past computer vision/object detection projects, ideally including at least one model taken to production or edge deployment.
• Solid understanding of mathematics relevant to computer vision (linear algebra, probability, optimization).
• Strong problem‑solving and analytical skills, with the ability to independently judge project feasibility, effort, and risk.
• Good communication skills, able to explain technical trade‑offs (offline vs. cloud‑based) to non‑technical stakeholders.
• Comfortable working cross‑functionally with product and engineering teams.
• Willing to work on-site/hybrid as needed for hardware testing and edge‑device deployment (adjust based on company policyn)