Back to all projects
AI / LLM
ITMO · 2018 – 2020

ML-Powered Pipeline Monitoring Platform

Real-time sensor classification platform that improved pipeline condition decisions by 20% over manual monitoring.

The problem

Field engineers relied on manual sensor readings to judge pipeline conditions (water saturation vs. oil readiness), which was slow and error-prone.

Impact

Decision-making accuracy
+20% vs. manual
Response time
Sub-second at scale

Approach

  1. 1

    Built a Flask-based platform ingesting real-time sensor data with a NumPy/Pandas feature engineering pipeline.

  2. 2

    Trained classification models with scikit-learn and TensorFlow to predict pipeline state in real time.

  3. 3

    Deployed with Docker on AWS EC2/S3 with Jenkins/GitLab CI for consistent, automated releases.

  4. 4

    Delivered training and documentation so field technicians could operate the system independently.

Stack

  • Flask
  • scikit-learn
  • TensorFlow
  • NLTK
  • NumPy
  • Pandas
  • Docker
  • AWS