How Greenlake helped an energy firm create a remote safety monitoring system


EnergyCorp is one of the first fully digital electricity providers in EU. Their smart energy meters enable their clients to have fine-grained information about their energy consumption patterns. Data which the smart meters provide help EnergyCorp create further value added services that delight their customers.


EnergyCorp wished to develop a non-intrusive solution for their senior citizen clients who were living independently, to enable their guardians to be warned in case of accidents or mishaps at home. Being sensitive and respectful of the wishes of their clients to live independent lives - intrusive technologies like wearable sensors were not an option. EnergyCorp hit upon the potential for their smart meter data to be used to flag mishaps at home.

After being unable to make any headway towards a solution to this problem using IBM's Watson, EnergyCorp turned to Greenlake for assistance in devising a solution to the challenge while simultaneously helping them grow the skills of their data science team.


In order to meet the client objectives, the project was partitioned into three stages

Stage 1: Determining the value proposition

As the commercial product was still being conceptualized, the engagement team started with a determining the target metrics expected from the product. These metrics were

  • The time to flag issues once they had occurred
  • A second, more technical, metric that ensured a balance between not missing any potential accidents, while avoiding too many false warnings.

Value created: Greenlake determined the suitable technical parameters required to create a solution enabling the client to focus on the needs of their customers.

Stage 2: Leveraging domain knowledge to focus further efforts

A rapid focussed analysis performed by the engagement team in order to select the most promising machine learning approaches for the problem.

Value created: This phase leveraged the extensive domain knowledge at Greenlake to dramatically reduce the time to arrive at a solution.

Stage 3: Knowledge transfer via hands on collaboration

This stage took the form of 3 months of supervision and guidance of the client's data science team in order to formulate and test the algorithms selected in Stage 2, in order to generate a solution that met the desired metrics obtained in Stage 1.

Value created:

  • Discussions, whiteboarding sessions and design sessions helped the client data science team to learn how to make efficient architectural decisions subject to the underlying tradeoffs.
  • Code reviews helped ensure that the implementation of the algorithms by the data science team led code that could be scaled and put into production.


The engagement yielded the following outcomes

  • A solution was architected and implemented in a form that reduced the time to detection of an incident by over 50% of the stated goals of the project!
  • Knowledge transfer during the project helped eliminate the need for a lengthy transitioning of the solution back to the client team. This reduced the time, friction and costs associated with such knowledge transfers.
  • Enabled a sense of ownership of the solution from the client data science team, enabling faster continued development of the project.
  • The client data-science team gained skills to tackle other projects of a similar nature - accelerating their task relevant maturity
  • The client data science team was made aware of pitfalls and common issues in data science projects - knowledge that help them avoid significant sunk costs.

Based on these successful outcomes, EnergyCorp commenced a followup phase of roll out of the solution to a large test group.

Have a similar problem?

If your firm is focused on the IoT-tech space space, let's discuss opportunities to strategically leverage machine learning to your problems.