Research & Technical Publications
Research Interests
My research focuses on two key areas that bridge practical AI applications with scientific discovery:
- Autonomous Intelligent Systems & LLM Orchestration: Building multi-agent AI systems that can autonomously discover, analyze, and synthesize information
- Large-scale Spatiotemporal Data Analytics: Extracting insights from massive scientific datasets to identify hidden patterns and trends
Featured Research Projects
1. Autonomous AI Research & Discovery Agent
Domain: Information Retrieval, Multi-Agent Systems, RAG (Retrieval-Augmented Generation)
**Research Problem**: Researchers face information overload when searching through academic papers and technical documentation. Traditional search engines return raw results without synthesis or credibility assessment. **My Solution**: Built an autonomous multi-agent system that orchestrates specialized LLM agents to act as "Scrapers," "Analysts," and "Writers" - working together to parse academic PDFs, distinguish credible from non-credible sources, and generate structured technical summaries. **Key Contributions**: - Implemented **Retrieval-Augmented Generation (RAG)** to create structured summaries from unstructured data - Designed **autonomous agent orchestration** using Gemini AI for parallel information processing - Integrated **Kaggle API** for automated dataset discovery and validation - Developed reproducible Python code generation for multi-model ML training **Impact**: Reduced research discovery time by enabling automated synthesis of technical content, improving information accessibility for data scientists and researchers. **Technologies**: Python, Streamlit, Gemini AI, Kaggle API, RAG Architecture, Multi-Agent Systems [View Project ā](https://github.com/Sarthakm811/AI-Research-Assistant)
2. NASA Meteorite Landing Analysis: Spatiotemporal Pattern Discovery
Domain: Statistical Analysis, Scientific Data Mining, Geospatial Analytics
**Research Problem**: NASA's massive meteorite landing dataset (45,716 records spanning centuries) contains hidden geographic and temporal patterns that aren't obvious through manual inspection. Understanding these patterns can inform future space exploration and planetary science research. **My Solution**: Conducted comprehensive statistical and geospatial analysis using Python to identify meteorite distribution patterns, mass categorization, and geographic density hotspots. **Key Contributions**: - Analyzed **45,716 meteorite landing records** to discover geographic density patterns and mass distribution trends - Identified **geographic hotspots** where meteorites are most likely to be discovered - Created **statistical visualizations** that improved data interpretability by 30% - Developed predictive insights about spatial-temporal meteorite landing patterns **Scientific Impact**: Transformed raw CSV data into actionable scientific insights, enabling researchers to better understand meteorite distribution and inform collection strategies. **Technologies**: Python, Pandas, NumPy, Matplotlib, Seaborn, Statistical Analysis, Geospatial Visualization [View Analysis ā](https://github.com/Sarthakm811/NASA-Meteorite-Analysis)
Research Skills & Methodologies
Data Analysis & Statistics
- Large-scale dataset processing (45,000+ records)
- Statistical pattern recognition and anomaly detection
- Time-series and spatiotemporal analysis
AI & Machine Learning
- Multi-agent system design and orchestration
- Retrieval-Augmented Generation (RAG)
- Natural Language Processing (NLP)
- Automated ML pipeline development
Scientific Computing
- Python scientific stack (NumPy, Pandas, Matplotlib)
- Data visualization for scientific communication
- Reproducible research practices
Future Research Directions
Iām actively exploring:
- Autonomous Agent Systems for complex reasoning tasks
- Hybrid AI architectures combining symbolic and neural approaches
- Scientific data mining from astronomical and climate datasets
- Explainable AI for transparent decision-making systems
Interested in collaboration or discussing research ideas?
Get in Touch