Ishaan Mishra
Selected work
Pairs a SWIN Transformer with CNN and CLIP backbones to tell real faces from AI-generated ones. The best hybrid reached 93.47% accuracy, and every hybrid beat its standalone backbone. Team project; paper in preparation.
Co-authored a comparative study of modern CNN and transformer models for classifying brain MRI scans as tumor or non-tumor. Presented at ICTIS 2026 in Bangkok; to appear in Springer LNNS.
Leave management for small teams that warns when too many teammates would be off at once. Separate employee, manager and HR roles, enforced with Spring Security and JWT.
Experience
Trained in the Java full-stack track (Spring Boot, Angular, SQL).
Built a role-based banking app with JWT auth and guardian approval for high-value transfers as the final project.
Built an intranet plotting tool in React that parses six data file formats into interactive 2D and 3D visualizations (Plotly.js, Three.js, Cesium.js).
Built web apps for social-enterprise projects; part of the team that won the Early Stage category at Enactus Nationals 2024.
Education
Now
Learning Japanese and going deeper into deepfake-detection research.