Amulya Kundalia

About

Computer science undergraduate at Delhi Technological University with research experience in deep learning and hands-on project work in full-stack development and machine learning. Worked across model development, reproducible experimentation, and backend/frontend implementation.

Experience

• Research Intern – Deep Learning
  Delhi Technological University (DTU) · New Delhi, India
  Jun 2026 – Jul 2026

  Led a 3-member research team developing TCEM, a generative deep-learning framework for learning biologically meaningful representations of breast-cancer gene expression. Architected and implemented a variational autoencoder with conditional decoding and distribution-alignment objectives across 11,172 genes and 1,195 TCGA-BRCA specimens. Designed a reproducible validation framework across 3 deterministic seeds, incorporating source-classification probes, mutual-information estimation, latent-space diagnostics, and downstream prediction tasks. Evaluated model generalization across 3 independent cohorts comprising 3,683 expression profiles, identifying and analyzing cohort-specific failure modes and representation shifts. Led experiment planning, debugging, code reviews, reproducibility checks, result interpretation, and technical documentation while co-authoring a research manuscript.

Education

• B.Tech. in Computer Science and Engineering
  Delhi Technological University
  Computer Science · Aug 2024

Skills

Tools / apps / platforms

Languages

Projects

• Social Media Application
  FastAPI, React.js, TypeScript, SQLAlchemy, SQLite, Tailwind CSS

  Built a full-stack social platform supporting JWT authentication, user accounts, image/video uploads, feeds, and owner-restricted content operations. Developed an asynchronous FastAPI and SQLAlchemy backend with relational user/post models, UUID-based records, and persistent database operations. Implemented core media/feed REST endpoints and integrated authentication and user-management route groups for registration, login, verification, password reset, and account management. Built a responsive React and TypeScript frontend using TanStack Query for asynchronous server-state management, API communication, and caching.

• KinetiQ
  FastAPI, PostgreSQL, Supabase, Python, Pandas, XGBoost, SHAP

  Engineered a secure FastAPI backend for managing athlete profiles, workload measurements, kinematics records, and injury-risk predictions. Designed a PostgreSQL/Supabase-backed application architecture supporting structured data ingestion, transformation, retrieval, and downstream prediction workflows. Built an XGBoost injury-risk pipeline with Python and Pandas preprocessing and integrated SHAP explanations for interpretable model predictions. Developed REST APIs with OAuth2 and JWT-based authentication for protected data access, prediction generation, and retrieval of application results.

Courses & certifications

• Unsupervised Learning · DeepLearning.AI | Coursera
• Supervised Machine Learning · DeepLearning.AI | Coursera
• Advanced Learning Algorithms · DeepLearning.AI | Coursera
• Neural Networks and Deep Learning · DeepLearning.AI | Coursera