About us:

VSOL is a digital enabler with a mission to help public and private organizations evolve their businesses through data and technology. We provide an end-to-end service from consulting to execution that drives the growth and innovation of our clients. As VSOL is in a phase of rapid expansion, we offer a dynamic, creative environment that accelerates your personal and professional development.

We are looking for talented individuals eager to develop in international markets while contributing to the company’s future in a constructive and supportive manner.

Role Overview:

We are looking for a Senior AI Engineer with strong machine learning and deep learning fundamentals and proven depth in computer vision. This position owns vision features end to end — from problem framing and data strategy through training, evaluation, and production deployment — and sets technical direction for the vision stack.

We value engineers who measure before they optimize, who treat data and label quality as part of the model, and who can explain the tradeoff behind a decision rather than only the result.

Responsibilities

Model Development

• Design, train, and optimize deep learning models for computer vision tasks, with primary focus on object detection and secondary focus on OCR and document understanding.
• Select architectures, loss functions, and training strategies based on accuracy targets, latency budgets, and available data rather than defaulting to the newest published model.
• Build and maintain reproducible training pipelines, including experiment tracking, configuration management, and dataset versioning.
• Apply classical computer vision and traditional machine learning techniques where they outperform deep learning in cost, latency, or auditability.

Data and Evaluation

• Own dataset strategy, including annotation guidelines, label quality audits, split design, and decisions about where annotation budget is spent.
• Define evaluation protocols that reflect production conditions, including per-slice metrics, operating-point analysis, and frozen regression sets.
• Diagnose model failures to root cause — data, labels, architecture, or pipeline — and convert each diagnosis into a repeatable check.

Production and Deployment

• Deliver models to production, including runtime export, quantization, and accuracy re-validation after conversion.
• Meet defined latency, throughput, and cost targets on the target hardware, whether edge devices or cloud GPU services.
• Implement monitoring for drift and silent degradation where ground-truth labels are unavailable, and define retraining triggers and release gates.
• Work with backend and platform engineers to integrate models into services, pipelines, and internal systems.

Collaboration and Technical Leadership

• Translate business requirements into technical specifications, including proposing a non-machine-learning baseline when that is the better answer.
• Review the work of other engineers, mentor junior and mid-level members, and raise the team’s standards for evaluation and reproducibility.
• Document architecture decisions, experiment results, and known limitations so the work can be maintained by others.
• Communicate results, risks, and tradeoffs clearly to product owners and non-specialist stakeholders.

Requirements

• Bachelor’s degree or higher in Computer Science, Electrical Engineering, Applied Mathematics, or a related field.
• Minimum 5 years of professional experience in machine learning or deep learning, including at least 3 years focused on computer vision.
• Demonstrated ownership of computer vision models deployed to production, not only proof-of-concept work.
• Strong command of object detection: architectures, label assignment, loss functions, evaluation metrics, and their failure modes.
• Practical experience with OCR or document understanding, including text detection, recognition, and output validation.
• Proficiency in Python and PyTorch or TensorFlow, with the ability to write production-quality, testable code.
• Solid foundation in machine learning fundamentals, including regularization, cross-validation strategy, class imbalance, data leakage, and model calibration.
• Experience with model optimization and deployment, including quantization, runtime export, and profiling against a latency or cost budget.
• Strong analytical and problem-solving skills.
• Good English communication skills and ability to work in an international team environment.

Nice to have: 

• Experience with Vietnamese-language OCR, including diacritics handling and Unicode normalization.
• Experience deploying to edge hardware (NVIDIA Jetson) or GPU serving platforms (Triton).
• Familiarity with vision-language models and a clear view of when they are appropriate versus task-specific models.
• Experience building or operating annotation pipelines or MLOps infrastructure.
• Contributions to open-source computer vision projects or published research.
• Experience mentoring engineers or leading a small technical team.

Why you’ll love working here

• Working in start-up environment, English-speaking, with opportunity to be part of innovation team and global projects.
• 13th-month salary, performance bonus.
• Premium Health insurance for employees and family members (depending on level), Annual Health Check, Government Insurance in probation.
• 14++ days of Annual leave and 5 days of Outing leave.
• Lunch allowance.
• Taxi & phone allowance (depending on level).