aishwarya kushwaha

I’m a B.Tech Computer Science student at Indira Gandhi Delhi Technical University for Women (IGDTUW), building projects across core CS and practical AI systems. My coursework includes Data Structures, Algorithms, Databases, Software Engineering, and Computer Networks.

In my research internship at IGDTUW (Dr. Nazme Zehra Naqvi), I designed and implemented an Attention U-Net++ and PatchGAN-based GAN architecture for latent fingerprint reconstruction. I built an end-to-end restoration pipeline using CLAHE, normalization, graph-based minutiae filtering, self-attention, and CBAM, aiming to preserve ridge continuity for noisy forensic inputs.

Beyond modeling, I evaluated performance across NIST SD302E/SD302G and FVC2002/2004, reporting strong biometric matching results with BOZORTH3 scores up to 75.03 and EER as low as 0.09. I care about building systems that generalize across datasets—not just demonstrations.

I also develop end-to-end AI products, like a meeting assistant that automates YouTube/local recording analysis using Whisper transcription, LLM summarization, and a RAG pipeline with LangChain and ChromaDB (Google Gemini API), powered through a configurable chunking workflow. I’m equally excited about shipping useful software, such as my full-stack movie app with React and Appwrite, including performance improvements like request optimization and efficient API caching.