Search

Knowledge transfer: how we support local autonomy

When working on a project in a specific region or country, we consider it a top priority to collaborate with local partners in a way that builds capacity. According to the United Nations, capacity building is the process of developing and strengthening skills and resources, among other abilities, in an organisation so that it can thrive and adapt in a fast-changing world.


Whether it’s a project to detect TB or some other infectious illness, the organisational skills are assessed, and we build capacity in various areas of data science, artificial intelligence and disease modelling. Through a collaborative approach, we use an agile four-level model, which is implemented during the project set-up and thereafter. Although this approach demands a serious commitment, program partners define the level of engagement and collaboration in view of capacity, timing and desired long-term strategy. Ultimately, it is in the partner and country's benefit to run the platform as autonomously as possible.


EPCON's AI model

The four levels include the following:


Level one:

  • Skills required: Knowledge of spreadsheets - e.g. Excel, Google Sheets and CSV formats; public health knowledge; M&E (monitoring and evaluation) know-how.

  • Activities performed: Explore what data can be used and where it will come from; collaborate with EPCON team on M&E and reporting requirements; engage with stakeholders.

  • EPCON remote support: Data cleansing (i.e. preparing data for analysis); model training and processing pipelines to ensure continuous data collection; Geoportal and dashboard visualisations (how the dashboard looks); core engine maintenance; technical and data science support.


Level two:

  • Skills required: Level one knowledge plus SQL programming; PostgreSQL (relational database management system); PostGIS (allows location queries to be run in SQL).

  • Activities performed: Create own data sets and training variables - we prepare the initial dashboard, but partners manage their own parameters; prepare data (do own data cleansing); analyse output and dashboard visualisations.

  • EPCON remote support: Technical and data science assistance; model training and processing data pipelines; geoportal visualisation; core engine maintenance and support.


Level three:

  • Skills required: Level one and two knowledge plus Python programming; SOM creation; basic Machine Learning.

  • Activities performed: Create new input variables - for example, the client has an HIV dataset and has the capacity to format it for the platform; add new disease profiles and co-morbidity; data preparation for training the model; analysing output and dashboard changes.

  • EPCON remote support: Technical and data science expert assistance; model training and processing data pipelines; core engine maintenance and support.


Level four:

  • Skills required: Level one, two and three knowledge plus R (for statistical computing); Nifi (for automating data flow between software systems); Elasticsearch (a search and analytics engine).

  • Activities performed: Full autonomy on data input variables; model training and output. Employees are almost like remote EPCON workers.

  • EPCON remote support: Technical and data science expert assistance; core engine maintenance and support.


EPCON’s knowledge transfer system incorporates various features that make it extra flexible for partners, including:

  • Evolving to partners’ goal level over time - this can be a gradual process.

  • Accelerating to higher levels, such as level three or four, if there are experienced software engineers on the team.

  • Shifting between levels - partners can fall back a level if, for example, they want to add a disease to the model, or require more expert advice and assistance


Many EPCON clients are not machine learning experts, but are public health specialists. EPCON’s framework of building local knowledge allows health providers to boost their capacities in a rigorous fashion.