# TechSodh — AI Research Agent

> TechSodh (https://techsodh.in) is an academic AI research assistant and literature discovery agent built in Bharat. It enables researchers, doctoral candidates, engineers, and scholars to perform semantic literature reviews on arXiv, ingest and index scientific PDFs using Retrieval-Augmented Generation (RAG), synthesize comparative research matrices, and query academic work across Indian languages and English.

## Primary Capabilities
- **arXiv & Open Web Academic Search**: Discovers peer-reviewed research papers, preprints, and open-access academic literature via semantic queries and field constraints (author, title, arXiv categories like cs.AI, cs.LG, math, physics).
- **Structure-Aware PDF RAG Ingestion**: Ingests multi-column scientific PDFs, extracts equations, section boundaries, and theorems with recursive sliding-window chunking and dense vector indexing.
- **Citation Attribution & Hallucination Mitigation**: Every generated claim is mapped back to exact passage offsets with direct arXiv e-print IDs, DOIs, and page references.
- **Multilingual Indic Support**: Cross-lingual semantic retrieval and synthesis across Indian languages including Hindi (हिन्दी), Tamil (தமிழ்), Bengali (বাংলা), Telugu (తెలుగు), Marathi (मराठी), and Sanskrit.
- **Persistent Multi-Session Workspaces**: Maintain multiple research threads, thesis topics, and annotated reading lists across sessions with persistent memory.

## Key URLs
- Homepage: https://techsodh.in/
- Research Workspace (App): https://techsodh.in/app
- Academic Research Documentation & Guides: https://techsodh.in/docs
- Privacy Policy: https://techsodh.in/privacy
- Terms of Service: https://techsodh.in/terms
- Contact: https://techsodh.in/contact

## Target Users
- Ph.D. scholars, Master's students, and academic researchers conducting systematic literature reviews.
- Machine learning, AI, and computer science practitioners tracking state-of-the-art preprint benchmarks.
- Multilingual and non-native English scholars reading and synthesizing dense English papers in vernacular languages.