Designed and built a multi-agent AI system that autonomously generates structured academic literature reviews from a single research topic by orchestrating specialized agents for paper discovery, PDF extraction, summarization, cross-paper comparison, and final review generation.

Implemented a backend pipeline to retrieve research papers from the web, extract text from PDFs, and generate structured analyses including methods, results, limitations, and research gaps.

Developed a full-stack research interface enabling users to generate publishable Markdown and LaTeX literature reviews; containerized the system using Docker for reproducible deployment.

Engineered a multi-agent orchestration pipeline integrating specialized parser, retrieval, reasoning, and reporting agents to automate end-to-end Anti-Money Laundering (AML) case investigation workflows.

Designed a modular data-processing architecture leveraging structured metadata propagation, multi-source information retrieval, and LLM-driven contextual reasoning to generate explainable risk assessments, suspicion scores, and investigator-grade AML narratives.

Implemented scalable agent communication with standardized JSON interfaces, asynchronous workflow orchestration, and robust validation pipelines, reducing average case processing time from 25–30 minutes to approximately 1.4 minutes while maintaining consistent and interpretable outputs

Built a backend data-processing pipeline to handle scanned answer sheets by extracting, structuring, and storing textual data for evaluation workflows.

Implemented an OCR-based preprocessing module to process image uploads and normalize extracted text, reducing manual correction effort by 30%.

Designed modular services to support file ingestion, processing, and result generation, improving reliability and maintainability of the system.