Three production AI systems, one long field-proven partnership.
Kongsberg Digital, now Falkor, brought XFactr in to bring order to a problem every sucker rod pump operator knows: too many dynacards, too few hours, and diagnostics that couldn't scale past a handful of SME reviewers. This is the real, honest account of what we built, how we built it, and what it changed.
REDUCED
A long transformation journey, not a one-off project
Falkor, a KONGSBERG company, is an industrial software business serving energy, maritime and utilities operators worldwide. The relationship with XFactr didn't start with a finished product brief. It started the way most real engagements do: with engineers embedded on the team, learning the domain, and earning the next stage of trust one delivery at a time. Here's the shape that journey actually took.
Embedded T&M Engineering
XFactr engineers joined as extended team members across Python, ML, GenAI, full-stack, data and QA, working inside Kongsberg Digital's own delivery cadence. See how our custom AI & ML development teams staff this way.
Full-Ownership Use Cases
As trust grew, XFactr took end-to-end ownership of defined problems: sucker rod pump diagnostics, well optimization, and CML coordinate extraction, each scoped, built and shipped by our team.
Research-Driven Proof of Concept
Every use case started as a scoped feasibility build, tested against real dynacards and real well data, before a single line of production code was committed. Backed by AI consulting & strategy.
Production Deployment
Containerized, monitored, and running in Kongsberg Digital's field and cloud environments today, with SME feedback loops feeding continuous retraining.
Kongsberg Digital is now Falkor, a KONGSBERG company
Falkor builds industrial software for the energy, maritime, chemicals and utilities operators who keep the physical world running, part of the 200-year-old KONGSBERG group. The oil and gas work below sits inside that mission: turning field data that used to live in a reviewer's head into something a whole operation can act on. This is the kind of long-horizon, deep-domain partner XFactr's enterprise AI applications practice is built around.
Three use cases. One field problem each.
Every tab below is a real, deployed system, not a lab demo. Built with our custom AI & ML development and data engineering teams working side by side.
Sucker rod pump diagnostics, from dynacard to decision in seconds
DEPLOYED · IN PRODUCTIONBUSINESS NEED
- SMEs spent significant time manually reviewing dynacards and correcting misclassified pump conditions
- No unified tool surfaced probabilities, severity levels, and key metrics like PPRL, MPRL and Fo together
- The system had to learn from SME feedback and improve over time
- Needed to be scalable, portable, and containerized for field or cloud environments
OUR SOLUTION
- Dual-input diagnostic system accepting either dynacard images or coordinate data
- Independent image and coordinate pipelines classify pump conditions in parallel
- UI surfaces Top-N predicted faults with severity, confidence, and calculated pump metrics
- SME feedback loop marks predictions correct or wrong, enabling continuous learning and retraining
VALUE CREATED
- Automates dynacard interpretation, cutting manual review load on field experts
- Standardized multi-label fault detection minimizes human subjectivity
- Engineers see probabilities, severity and metrics to prioritize action faster
- Fully containerized, deployable across wells, edge devices, or the cloud
Well optimization: early warning across an entire fleet
853 WELLS · 4 PUMP TYPESBUSINESS NEED
- Operators lacked early visibility into well failures, so interventions stayed reactive and expensive
- Event and sensor data was fragmented across disconnected operational systems
- Historical failure labels were incomplete and retrospectively logged, making training difficult
OUR SOLUTION
- End-to-end data engineering and ML platform ingesting and reconciling production and sensor data
- Multi-stage label correction pipeline that identifies true event boundaries from production signal
- Two-stage predictive architecture: anomaly detection for early warning, then fault classification
VALUE CREATED
- Advance warning ahead of failures lets operations teams plan instead of react
- Automated fault classification narrows root cause from broad event to specific sub-category
- Tiered imputation extracts maximum signal from partially instrumented fleets
- Pipeline adapts to new fields, pump types and event taxonomies without rebuilding
Built on the same data platforms & lakehouse and MLOps & DevOps foundations we bring to every field-scale deployment, running on cloud infrastructure sized for hundreds of wells at once.
Automated CML coordinate extraction from piping ISO drawings
COMPUTER VISION · NO LLM DEPENDENCYBUSINESS NEED
- Manual extraction of CML coordinates from piping ISO drawings was slow and error-prone
- Complex drawing styles, colored leader lines, and inconsistent OCR labels made detection difficult
- Engineers needed structured JSON/CSV output without relying on LLM-based processing
OUR SOLUTION
- Computer vision and OCR pipeline using OpenCV and skimage to detect CML bubble types
- Colour masking, skeletonization, and leader-line tracing identify pipe attachment points
- Multi-stage workflow extracts anchor coordinates, elevation labels, pipe topology, and direction
VALUE CREATED
- Detects CML tags and extracts coordinates directly from ISO drawings with minimal manual effort
- Leader-line tracing and skeletonization deliver reliable coordinate detection
- Handles complex drawing styles, colored annotations, and varied PDF layouts
- Outputs (N, E, EL) coordinates as structured JSON/CSV, ready for inspection platforms
What changed once these systems went live
These are the numbers from the Dynacard automation engagement specifically, the same platform Falkor's field engineers rely on today to focus their time on exceptions instead of every single reading.
One team, deployed across every layer of the stack
This engagement ran on XFactr's time-and-materials model: engineers embedded directly into Kongsberg Digital's delivery team, spanning data science, GenAI, full-stack engineering, and quality assurance, not a single specialist handed a narrow brief. Here is the real breadth of skills our engineers brought to the table.
PROGRAMMING & CORE
Python, SQL, JavaScript, Go, C, PHP, PL/SQL, Java, microservices and REST/gRPC API design.
AI / ML
Machine learning, deep learning, NLP, computer vision, time series and predictive analytics, MLOps. See custom AI & ML development .
GENAI / LLM
LangChain, LangGraph, RAG, multi-agent systems, LLM fine-tuning, prompt engineering, vector databases. See our generative AI services and agentic AI solutions .
FRAMEWORKS & TOOLS
PyTorch, TensorFlow, Keras, Hugging Face, FastAPI, Streamlit, OpenCV, YOLO, scikit-learn.
DATA / BI
Power BI, Tableau, Pandas, MySQL, PostgreSQL, MongoDB, ETL pipelines. See data & analytics and analytics & governance .
CLOUD & DEVOPS
AWS (Bedrock, SageMaker), Docker, Kubernetes, GitHub Actions, MLflow, CI/CD, Prometheus, Grafana. See MLOps & DevOps , AIOps solutions and cloud services & migration .
FULL-STACK ENGINEERING
ReactJS, Angular, React Native, Node.js, Django, distributed systems. See full stack development and APIs & microservices .
QUALITY ENGINEERING
Selenium, Cypress, WebDriverIO, API and regression testing, Agile/Scrum. See quality engineering and test automation .
More of what XFactr engineers, beyond this engagement
This case study touched AI, data, GenAI, cloud and QA. Here is the rest of what that same engineering discipline covers, across every layer we work in.
AI & DATA
AI X OPS & DIGITAL ENGINEERING
The same engineering, applied across the energy value chain
The field-data discipline behind this engagement, ingesting sensor and event data reliably, then acting on it in near real time, is the same discipline behind XFactr's wider energy and industrial practice.
ENERGY & UTILITIES
FIELD SYSTEMS
MONITORING & ANALYTICS
Long-term partnerships, not one-off projects
Kongsberg Digital, now Falkor, is one of a small number of clients XFactr partners with long-term, the same model behind our work with Schneider Electric and WESCO. Fewer clients, deeper focus, architect-led delivery every sprint.
Energy Management Solutions
See how this same engineering discipline applies to grid, microgrid and utility operators.
Enterprise AI Applications
The broader practice this engagement was built inside of, applied across manufacturing, buildings and more.
AI Consulting & Strategy
Where every one of these use cases began, as a scoped feasibility conversation.
What this page is built to be found for
The keyword targets below reflect real buyer intent for this case study: operators and engineering leaders searching for proof that AI works in oil and gas, not just marketing claims about it.
| KEYWORD | SEARCH INTENT | PRIORITY |
|---|---|---|
| oil and gas AI case study | Buyer researching proof, late-stage vendor evaluation | High |
| sucker rod pump diagnostics AI | Technical, solution-aware, high intent | High |
| dynacard analysis automation | Problem-aware, technical buyer | High |
| predictive maintenance oil and gas wells | Solution-aware, comparing vendors | High |
| well failure prediction machine learning | Technical, engineering-led search | Medium |
| computer vision piping ISO drawings | Highly specific, niche technical buyer | Medium |
| CML coordinate extraction automation | Highly specific, inspection/engineering teams | Low, high-value |
| AI system integrator for energy companies | Vendor search, buyer intent, comparison stage | High |
| Kongsberg Digital Falkor AI partner | Brand + partner search, high trust intent | Medium |
| time and materials AI engineering team | Staffing/engagement-model buyer intent | Medium |
| generative AI services for industrial companies | Solution-aware, exploring GenAI fit | High |
| MLOps for field deployment | Technical, engineering leadership | Medium |
| proof of concept to production AI | Decision-stage, evaluating delivery model | High |
| enterprise AI applications energy sector | Solution-aware, budget owner | High |
Bring us a field problem this real. We'll bring the engineering.
Fewer clients, deeper partnerships, architect-led delivery, the same model that took Falkor from embedded T&M engineering to three production AI systems.