Priyanshu Rawat
Building production RAG systems for financial AI
Overview
Data Scientist @FLX AI
Data Scientist @Center for Integrated Research Computing, UoR
Rochester, NY
he/him
Social Links
About
Data Scientist and AI Engineer with expertise in building production-grade RAG systems and agentic AI solutions. Currently developing financial document processing systems at FLX AI and HPC-integrated chatbots at University of Rochester's Center for Integrated Research Computing.
Background in computer vision research and distributed machine learning with strong foundation in PyTorch, TensorFlow, and cloud technologies. Passionate about creating AI solutions that solve real-world problems in finance, research computing, and enterprise automation.
One of my key projects, FinRAG3, is a 6-phase agentic AI system that processes SEC filings and fund prospectuses with 95% accuracy, reducing analysis time from hours to 5 minutes. I've also developed Steam Insights, a comprehensive gaming market analysis system processing 8M+ data points with ML forecasting.
Let's connect and collaborate on cutting-edge AI solutions!
Stack
Blog
Experience
FLX AI
Current Employer- Architected FinRAG3, a 6-phase agentic AI system using LangGraph orchestration for automated investment due diligence, processing SEC filings (10-K, 10-Q) and fund prospectuses with custom parsing algorithms achieving 95% accuracy in regulatory data extraction and reducing analysis time from hours to 5 minutes
- Developed domain-specific parsers for complex financial documents including SEC reports, fund prospectuses, and regulatory filings, implementing intelligent chunking strategies for risk factors, MD&A sections, and fee tables while achieving 5-10x processing speed improvements
- Built production RAG infrastructure handling 500+ page financial documents with multi-GPU optimization, ChromaDB vector storage, and comprehensive A/B testing framework, achieving >90% accuracy on investment questionnaire automation
- Deployed enterprise-grade MLOps pipeline using Docker, Kubernetes, and Flask-based monitoring dashboard, achieving 95% system reliability improvement while enabling real-time financial document analysis
- LangChain
- FastAPI
- ColBERT v2
- PostgreSQL
- Docker
- GPU Computing
- ChromaDB
- LangGraph
Center for Integrated Research Computing, UoR
Current Employer
K-Labs: Continual Learning Lab, University of Rochester
Greene Career Center, UoR
Insignia Consultancy
Education
University of Rochester
Rochester, New York
Key Coursework
- Machine Learning
- Computational Statistics
- Data Science at Scale
- End-to-End Deep Learning
Graphic Era Hill University
Dehradun, India
Key Coursework
- Machine Learning
- Data Structures and Algorithm
- Deep Learning
Projects(4)
Comprehensive gaming market analysis and forecasting system processing 8M+ data points from 140K+ games. Built ETL pipeline with Apache Airflow, Databricks Spark, and Kafka. Developed ML models (XGBoost, Random Forest) for review analysis and pricing forecasts. Implemented time series forecasting (ARIMA, Prophet) achieving 85% accuracy in genre demand predictions and reliable sales forecasting.
- Apache Airflow
- Databricks Spark
- Kafka
- XGBoost
- Random Forest
- ARIMA
- Prophet
- Python
Honors & Awards
Awards and achievements will be displayed here.
Academic and professional recognitions coming soon!