About Cynapse
Cynapse is a leading AI software company specializing in enterprise-grade Video Intelligence Solutions Powered by Generative AI, tailored to meet the unique challenges of various industries. Our vertical-specific solutions empower organizations to enhance safety, operational efficiency, and security in complex environments such as roads, seaports, airports, and cities. By combining advanced Vision AI with Generative AI, we continually push the boundaries of video analytics, delivering insights and automation that transform operations.
Led by a global team with a proven track record of scaling startups into market leaders, we foster innovation, collaboration, and diverse perspectives. Headquartered from US, Cynapse serves clients worldwide, redefining what's possible with video intelligence.
Job Description
We are seeking a talented and highly motivated Machine Learning Engineer to join our Model Engineering Team, focused on building, training, and improving computer vision models used in real-world systems. In this role, you will engineer models for improved accuracy, speed, and reliability, and work across experimentation, data, and production-facing ML pipelines.
You will collaborate with experienced engineers to translate research ideas and experiments into practical improvements that impact deployed systems.
You will also work with production-level machine learning systems, contributing to how models are evaluated, deployed, and monitored in real-world environments.
What You’ll Do:
Model Engineering & Training
- Assist in training, fine-tuning, and evaluating computer vision models for tasks such as classification, object detection, segmentation, pose estimation, and video understanding.
- Contribute to model engineering efforts, including modifying architectures, adjusting backbones or heads, and tuning training strategies to improve accuracy, stability, and inference speed.
- Experiment with training techniques such as data augmentation, loss functions, optimisation methods, and learning rate scheduling.
- Explore and evaluate modern vision models, including open-vocabulary or multimodal approaches where relevant.
Research & Experimentation
- Read, discuss, and apply insights from relevant research papers, technical blogs, and internal technical documentation.
- Prototype and evaluate new models, approaches, or techniques, comparing them against established baselines through structured experiments.
- Analyse experimental outcomes and document insights on performance trade-offs, limitations, and failure modes.
Evaluation & Failure Analysis
- Analyse model performance beyond headline metrics, including long-tail behaviour, edge cases, and real-world impact.
- Debug model failures such as false positives, false negatives, or unstable predictions.
- Improve robustness through targeted data improvements, inference logic, or model adjustments.
Data & Experimentation Infrastructure
- Work with large-scale datasets, including preprocessing, augmentation, validation, and quality checks.
- Follow established workflows for dataset versioning, experiment tracking, and reproducibility.
- Write clean, maintainable experiment code and configuration-driven training setups.
Production & MLOps Exposure
- Assist with model deployment and monitoring under the guidance of senior engineers.
- Work with training, evaluation, and inference pipelines used in production.
- Apply model lifecycle practices such as versioning, retraining, validation, and rollback.
- Consider performance and efficiency constraints, including latency, GPU memory usage, and inference cost.
Minimum Requirements
- At least a degree in Computer Science, AI, Machine Learning, or a related technical field, or equivalent practical experience.
- 1 – 2+ years of relevant hands-on experience with deep learning and computer vision through professional work, research, or substantial projects.
- Proficiency in Python.
- Experience with PyTorch and/or TensorFlow.
- Familiarity with Linux environments (CLI usage, basic shell commands).
Bonus Skills
- Exposure to multimodal or open-vocabulary models.
- Experience modifying models beyond configuration-level changes (e.g. layers, heads, loss functions).
- Familiarity with experiment tracking or MLOps tools (e.g. Git, Docker, MLflow, Weights & Biases).
- Experience with model optimisation or export (e.g. ONNX, TensorRT, quantization).
- Experience working with video data or tools such as OpenCV or FFmpeg.
Personal Competencies
- Strong interest in engineering and improving machine learning models, rather than treating them as black boxes.
- Curiosity about how architecture, data, and training strategies affect real-world performance.
- Comfortable experimenting, debugging, and iterating in a collaborative engineering environment.
- Ability to take ownership of model components appropriate to experience level and project needs.
Note: Due to the nature of the role, the position is considered only for candidates who are already based in Singapore.
