InsightAgent AI

Inspiration

Enterprise teams spend countless hours manually parsing tabular spreadsheets, identifying statistical anomalies, and extracting structural information from visual diagrams. InsightAgent AI was built as an autonomous multi-modal agent to combine machine learning data analytics with computer vision into a single unified interface.

What it does

InsightAgent AI operates across two core pipelines:

• Automated Data Analytics: Ingests CSV and Excel files to compute statistical summaries (standard deviation, variance, skewness) and runs an unsupervised Isolation Forest machine learning model to flag data anomalies on interactive Plotly dashboards.
• Vision AI Engine: Processes report diagrams and flowchart images using an OpenCV vision pipeline (adaptive thresholding and contour detection) to segment shapes, isolate structural boundaries, and extract region metrics.

How we built it

• Frontend: Streamlit with customized UI layout.
• Data & Machine Learning: Pandas, NumPy, Scikit-Learn (Isolation Forest algorithm), and SciPy.
• Computer Vision: OpenCV (cv2) and Pillow for contour detection and visual segmentation.
• Visualization: Plotly Express and Plotly Graph Objects.

Challenges we ran into

Integrating two distinct pipelines—unsupervised machine learning for tabular datasets alongside structural computer vision processing—into a single responsive UI without sacrificing execution speed or usability. Ensuring seamless execution across varying input file formats required careful error handling and robust backend architecture.

Accomplishments that we're proud of

Successfully building a fully functional, multi-modal application end-to-end as a solo developer within a tight timeframe. We are especially proud of creating a seamless user experience that allows users to switch effortlessly between tabular ML diagnostics and visual vision tasks.

What we learned

Building InsightAgent AI deepened our expertise in streamlining multi-modal workflows, optimizing OpenCV contour algorithms, and integrating machine learning diagnostics into practical web interfaces for non-technical end users.

What's next for InsightAgent AI

• Adding LLM-powered natural language executive summaries to explain detected anomalies automatically.
• Expanding computer vision capabilities with OCR text recognition for full diagram text extraction.
• Implementing one-click PDF diagnostic report exports.

Built With

• computer
• data-analytics
• machine-learning
• numpy
• opencv
• pandas
• plotly
• python
• scikit-learn
• streamlit