Lineage

home page
Pipeline visualization
Pipeline Visualization with segments and quick access to data chunks
Complete Dataset
Pipe Image Generator with Featherless.ai and Gemini Image
Pipeline visualized with nanobanana pro (mockup)
Initial Workflow / Tech Stack

Inspiration

RCP currently spends weeks manually aligning in-line inspection (ILI) datasets to understand how corrosion and other anomalies evolve. This slow, error-prone process delays repair decisions, increases operational risk, and creates compliance challenges. We were inspired to build Lineage to automate alignment, quantify corrosion growth, and shift pipeline integrity from reactive investigation to predictive, data-driven decision-making.

What it does

Lineage is an AI-powered pipeline integrity intelligence platform that ingests multiple ILI inspection runs, aligns reference points such as girth welds and valves, matches evolving anomalies across time, calculates corrosion growth rates, and highlights anomalous behavior. The system applies machine learning–driven (XGboost) similarity scoring and confidence estimation, then presents results through a clean, interactive dashboard for rapid engineering review. By reducing analysis time from weeks to hours, Lineage enables faster maintenance prioritization, improved safety, and stronger regulatory readiness.

How we built it

We designed Lineage as a full-stack AI application.

Backend: Python with FastAPI for scalable data ingestion, alignment logic, anomaly matching, and growth-rate computation.
Database: MongoDB Atlas to store large inspection datasets, anomaly metadata, and computed results while supporting flexible schema evolution.
Frontend: Next.js for responsive visualization, anomaly exploration, and integrity insights.
Machine Learning: scikit-learn models for anomaly detection, similarity scoring, and corrosion growth estimation across inspection runs.
AI Visualization: Google Gemini integrated with Nanobanana to generate clear, interpretable visual summaries and insights from processed inspection data.

Challenges we ran into

One of the biggest challenges was handling noisy, inconsistent real-world inspection data, missing values, unit mismatches, and positional drift between inspection runs, making anomaly alignment difficult.
We also faced technical hurdles integrating MongoDB for efficient storage and querying of large inspection datasets, requiring us to design flexible schemas and optimize performance.
Another challenge was ensuring our scikit-learn models produced meaningful anomaly and growth insights while remaining explainable for engineering use.
Finally, integrating Gemini-based visualization in a way that enhanced clarity, rather than adding ambiguity, required careful prompt design and validation.

Accomplishments that we're proud of

We delivered a working end-to-end platform that automatically aligns inspection runs, detects corrosion growth, flags anomalous behavior, and visualizes results in an intuitive interface.
Lineage demonstrates how machine learning and scalable cloud architecture can meaningfully reduce analysis time, improve defect detection accuracy, and support proactive integrity management for RCP.

What we learned

Through this project, we learned how to:

Apply scikit-learn machine learning models to detect anomalies and estimate corrosion growth over time.
Design and integrate MongoDB as a scalable data store for complex industrial inspection data.
Use Google Gemini with Nanobanana to generate clean, interpretable AI-assisted visualizations.
Build production-style full-stack data and AI systems that balance performance, usability, and explainability in safety-critical environments.

We also gained a deeper appreciation for the challenges of real industrial data and the importance of trustworthy, interpretable analytics in infrastructure applications.

What's next for Lineage

Next, we plan to expand Lineage with multi-run temporal modeling, improved growth-prediction accuracy, confidence-aware risk scoring, and deeper digital-twin integration for RCP’s assets.
Our long-term vision is a fully predictive pipeline integrity platform that continuously learns from inspection history, forecasts risk mathematically over time, and enables truly proactive infrastructure management.