Anushree Joshi

Engineered a multimodal deception-detection pipeline in PyTorch, fusing video (YOLOv8 face tracking, InsightFace, 3D pose) and audio (WavLM-Large diarization) features to reach AUC 0.71 for per-subject behavioural anomaly detection across 30+ clips. Validated the pipeline using leave-one-subject-out (LOSO) cross-validation and graph-autoencoder falsification across 134 feature channels. Deployed the model as a FastAPI service with CUDA and TensorRT-backed inference; presented findings to the DRDO journal board.

AI/ML Intern at Defence Institute of Psychological Research (DIPR), DRDO (2026-06 – 2026-07)

• Engineered a multimodal deception-detection pipeline in PyTorch, fusing video (YOLOv8 face tracking, InsightFace, 3D pose) and audio (WavLM-Large diarization) features to reach AUC 0.71 for per-subject behavioural anomaly detection across 30+ clips.
• Validated the pipeline using leave-one-subject-out (LOSO) cross-validation and graph-autoencoder falsification across 134 feature channels.
• Deployed the model as a FastAPI service with CUDA and TensorRT-backed inference; presented findings to the DRDO journal board.

AI-Powered Full-Stack Development Intern at IGDTUW (2025-06 – 2025-08)

• Developed and deployed a full-stack AI-integrated web application (React.js, Node.js, Express.js, MongoDB), contributing across backend logic, database design and frontend features.
• Integrated OpenAI and Hugging Face APIs for automated text processing, reducing manual effort by ~30%.
• Built authenticated REST APIs with request validation, optimised MongoDB queries to improve response times, and automated AWS deployment using GitHub Actions CI/CD.

B.Tech in Information Technology – Indira Gandhi Delhi Technical University for Women (2024-08)

Diploma in Computer Science and Engineering – Delhi Skill and Entrepreneurship University (2021-12 – 2024-06)