Research focus: multimodal deep learning for cancer prognosis in the Faculty of Medicine, University of Helsinki.

• Built an end-to-end multimodal pipeline integrating bone-marrow single-cell morphology and clinical variables.
• Encoded cells into 384-dimensional embeddings with a self-supervised vision transformer and trained a plasma-cell/plasmablast classifier.
• Developed a multiple-instance-learning survival model and fused image predictions with clinical Cox models.
• Submitted first-author conference abstract on multimodal survival prediction in newly diagnosed multiple myeloma.

Designed and built an AI academic-writing refinement SaaS for English and Chinese users.

• Implemented split-panel diff editing, discipline controls, intensity controls, and full LLM API integration.
• Used Next.js, Gemini 2.5 Flash API, Supabase, and Vercel for a commercial freemium product.
• Handled product design, implementation, deployment, and iteration independently.

Designed and developed a corporate website and brand mark for a Finnish granite-quarrying enterprise.

• Built a Vite React 18 TypeScript frontend with multilingual public pages, catalog pages, filters, search, and project galleries.
• Implemented a Node.js/Express backend with Zod validation, Multer uploads, Helmet middleware, admin audit history, and PostgreSQL-ready persistence.
• Deployed the frontend on Vercel and designed a premium SVG logo and visual identity for the company.

R-based visual analysis project focused on high-dimensional data and dimensionality-reduction workflows.

• Explored statistical visualization approaches for making high-dimensional structure easier to inspect.
• Applied R-based data analysis workflows for dimensionality reduction and visual interpretation.
• Maintained the project as a public GitHub repository for reproducible review.