Advika Sinha

Advika Sinha

Core Member, DSG IIT Roorkee

Education

  • B.Tech, Mechanical Engineering, IIT Roorkee (2023–2027)

Research Interests

  • SysML / Hardware aware ML
  • AutoML
  • Natural Language Processing
  • Deep Generative Models

Projects

  • GANalyse: Profiling & Optimization Framework for DL Architectures Developed a PyTorch Profiler-based analysis pipeline for architecture-specific optimization, using ResNet as baseline. Cross-verified profiler metrics with literature to automate failure detection and evidence-based tuning. (DSG, IIT Roorkee, 2024–Present)

  • 2D to 3D Modeling using Deep Learning Designed a deep learning pipeline to reconstruct 3D CAD models from 2D engineering drawings, integrating CNN-based shape estimation with SolidWorks API via VBA scripting. (MIED Department, IIT Roorkee, 2025)

  • Sirius: Multi-Retrieval RAG System for Financial and Legal Docs Developed AI agents using Corrective RAG (CRAG) and PlanRAG. Integrated HyDE to enhance dense zero-shot retrieval and evaluated multi-retrieval pipelines on financial QA tasks. (Inter IIT Tech Meet 13.0, IIT Bombay, 2024)

  • Knowledge Graph Embeddings for RAG LLMs Created a unified image-text knowledge graph using CLIP/USE embeddings and dimensionality reduction (UMAP, t-SNE, PCA). Integrated KGE to improve RAG model context. (DSG, IIT Roorkee, 2024)

  • StyleUp: GenZ Fashion Design Platform Built a creative design app using Neural Style Transfer and DALL·E to generate wearable art. Streamlit-based app ranked in top 4% at Myntra HackerRamp. (Myntra Hackerramp: WeForShe 2024)

  • Tech-Enhanced AI Interview Learning Platform Built an AI-driven platform generating context-specific interview questions using ML models. Integrated Google Speech Recognition to capture and assess spoken responses. Web-scraped dynamic datasets to deepen conversational relevance. (Techshila, Inter Bhawan Tech Meet, IIT Roorkee, 2024)

  • Dynamic Adaptation of Digital Workspace Designed an adaptive model using Random Forest and Windows API to optimize system resources based on user eye movement and activity. (DSG’s Bring Your Own Project, IIT Roorkee, 2023)

Publications

  • Puniani, C., Sinha, A., Singhi, S., & Yadav, A. (2025). Impact of Language Guidance: A Reproducibility Study. Under review at TMLR.

Roles at DSG

  • BYOP’25 Mentor
  • Part time developer

Follow

GitHub LinkedIn Medium

Updated: