Smart Digital Drilling Incident Investigator

AI-powered tool for upstream oil & gas teams to search, analyze, and learn from past drilling incidents using MongoDB vector search and Gemini AI for fast, intelligent safety insights

Comment

Inspiration

In the high-risk world of offshore oil & gas operations, safety isn't optional—it's mission-critical. We were inspired by the need to make incident data more actionable for engineers and safety teams. With thousands of past drilling incidents buried in reports, we asked: What if AI could help uncover patterns, prevent future failures, and save lives?

What it does

Smart Digital Drilling Incident Investigator is an AI-powered web app that allows users to:

  • Perform semantic search across historical incident reports.
  • Discover similar past incidents using MongoDB vector search.
  • Ask natural language questions and receive AI-generated insights from Gemini. It helps upstream engineers and safety analysts explore and understand offshore incidents more intelligently and intuitively.

How we built it

  • We sourced the BSEE Incidents of Non-Compliance (INC) dataset, cleaned and structured it for relevance.
  • Used Gemini to generate vector embeddings from incident descriptions.
  • Stored data and vectors in MongoDB Atlas with vector search indexing enabled.
  • Built a backend using FastAPI deployed on Google Cloud Run to handle search, Q&A, and similarity matching.
  • Designed a clean and intuitive React-based frontend hosted on Firebase to interact with the AI.

Challenges we ran into

  • Cleaning and standardizing messy real-world incident data across years and formats.
  • Managing and optimizing vector embedding sizes for effective similarity search in MongoDB.
  • Balancing response accuracy with latency in the AI-driven Q&A experience.
  • Integrating multiple cloud tools (MongoDB, Vertex AI, Firebase, Cloud Run) within the short timeframe.

Accomplishments that we're proud of

  • Seamlessly integrated vector search and generative AI in a meaningful industrial use case.
  • Created a fast, intuitive interface that lets users interact with years of complex safety data in seconds.
  • Delivered a full-stack AI solution within 6 days, ready to scale and extend to broader datasets.

What we learned

  • How to harness the power of MongoDB Atlas Vector Search for real-world semantic applications.
  • How to integrate Gemini AI into structured workflows using Retrieval-Augmented Generation (RAG).
  • The value of combining structured data, unstructured text, and AI to unlock deep industry insights.

What's next for Smart Digital Drilling Incident Investigator

We plan to expand the dataset to include global incident reports, integrate real-time safety alerts, and enable predictive analytics for early risk detection. Long-term, we envision this tool becoming a safety companion for every upstream operations team—turning hindsight into foresight with AI.

Built With

  • bsee-data-log
  • fastapi
  • firebase
  • google-cloud-run
  • google-gemini-api
  • mongodb-atlas
  • mongodb-atlas-vector-search
  • python
  • react.js
  • retrieval-augmented-generation
  • sentencetransformers
Share this project:

Updates