Data Scientist, Multimodal AI, AI Practice
Summary
Research and prototype multimodal AI models (vision, language, audio) for Singapore’s public sector, evaluating their fit for real-world government tasks.
Data Scientist, Multimodal AI, AI Practice at GovTech.
About the role Join the AI Practice team at GovTech to research and implement multimodal artificial intelligence solutions across Singapore government services. You will bridge the gap between emerging research and practical public sector applications by building prototypes and evaluating new model capabilities.
Key facts
Location: Singapore
Engagement: Full-time
Team: Multimodal AI Squad, Government Technology Office (GTO)
What you'll do
- Test and evaluate multimodal models including VLMs, MLLMs, diffusion, speech, and audio systems for government strategy.
- Create and execute benchmarks to assess AI model performance on real-world government tasks.
- Develop functional proofs-of-concept for areas such as smart city video analytics, document intelligence, generative media, and content authentication.
- Monitor research developments to determine which technologies are suitable for government deployment.
- Maintain internal prototyping platforms and multimodal AI showcases.
- Author technical reports, handbooks, and evaluation summaries for internal stakeholders.
Requirements
- Degree in Computer Science, Data Science, AI/ML, Engineering, or a related discipline.
- Experience with Python and deep learning frameworks, specifically PyTorch.
- Knowledge of machine learning fundamentals, including experiment design, model training, and evaluation methods.
- Understanding of computer vision concepts or direct experience with image, video, or audio data.
- Familiarity with or interest in multimodal AI, vision-language models, and generative AI.
- Ability to explain technical concepts to non-technical agency stakeholders.
- Comfort working in ambiguous environments where problem scopes are not strictly defined.
Nice to have
- Proficiency in using AI-assisted development tools like Cursor, Claude Code, or OpenAI Codex to improve coding speed and debugging.
Skills & tools
- Python
- PyTorch
- Multimodal AI (Vision, Language, Audio, Video)
- Generative AI
- Model Evaluation & Benchmarking
- Prototyping
Practical notes
- By submitting an application, you consent to GovTech and relevant government agencies verifying your information with third parties.
- Personal data provided may be shared across government agencies for recruitment and reporting purposes.
- All application information must be accurate; providing false information may result in the rejection of your application or termination of employment.