Case study
Company details
Offices:
Poland
Company size:
60-200 employees (medium)
Industry:
Government
Revenue:
Non-profit org
NO TIME TO WASTE
Our goal was to create dashboard providing
data-driven insights for station dispatchers
Exploratory analysis of the received data, projecting data-flow, identification of additional data sources and crystallization of the technical implementation of the business process
Building predictive models for inference values every 10min in span of 24h. For the purpose of PoC implementation, the problem has been simplified.
We build a production-ready solution, based on LLM models. Our ML team together with Surfer's frontend and backend specialists focused on creating a release version of the product. A well-evaluated pipeline of more than 30 distinctive ML tasks was created with 3 major technological breakthroughs in NLP released during development. About a week before the planned release, OpenAI released a game-changing GPT4-32k model that we managed to integrate & evaluate to the final solution before launch. Thanks to us, Surfer was the first company in the SEO market with this kind of technology inside.
Absorb domain knowledge
The field of urban water management is an extremely specific field with its own unique data characteristics. Part of our work was to learn about the nature of the processes and extract knowledge from experts
Enriching time series prediction with additional variables
The values of pressure and flow of water are continuously read by sensors monitoring the city's water supply, thus we can consider the problem of predicting successive values as a time-series prediction. In order to improve the quality of the model, it was also provided with external data,. which may affect the city's water resources management such as the weather or ongoing mass events.
Conformal prediction
Not only did we provide detailed information about water supply indicators but also showed how certain the model is. Confidence intervals were added to charts allowing quick identification of areas where the model lacked confidence.
Anomaly studies
The system for delivering water to residents is crucial, and absolutely no one can afford failures. Therefore, we have built a system that detects detections based on time-series data.
A few thoughts from our co-founder