©2023 Paul Zana
Web App for Air Pollution Device Testing
(B2B, AI/ML)
(B2B, AI/ML)
I led the design of an intuitive web app that empowers AirQo's hardware team and other organizations to closely monitor the air pollution data collected by AirQo devices. The tool seamlessly integrates machine learning algorithms(AI) — eliminating any reliance on having an on-site data scientist for performance reports, ensuring accurate device approval and quality device manufacturing.
Problem statement
AirQo, a leading player in air quality monitoring, faced a critical challenge in its hardware approval process. The existing process relied heavily on manual reports generated by data scientists, resulting in considerable delays and inaccuracies in approving devices for deployment on the AirQo network. On average, it took approximately one week to create and review these reports, impeding the timely and accurate approval of devices.
AirQo, a leading player in air quality monitoring, faced a critical challenge in its hardware approval process. The existing process relied heavily on manual reports generated by data scientists, resulting in considerable delays and inaccuracies in approving devices for deployment on the AirQo network. On average, it took approximately one week to create and review these reports, impeding the timely and accurate approval of devices.
Challenge
The primary challenge lay in the hardware team's inability to effectively analyze the machine learning (ML) datasets required for device approval without the direct involvement of data scientists. This dependency on data scientists significantly slowed down the device approval and deployment processes, impacting AirQo's ability to scale its operations efficiently.
The primary challenge lay in the hardware team's inability to effectively analyze the machine learning (ML) datasets required for device approval without the direct involvement of data scientists. This dependency on data scientists significantly slowed down the device approval and deployment processes, impacting AirQo's ability to scale its operations efficiently.
Solution
As a lead designer, I proposed and implemented an innovative solution to address these challenges. Leveraging the power of AI/ML, I designed a web application that empowered the hardware team with the autonomy to test and approve devices for deployment.
Success Metrics
1. Time on Report and Approval
2. Hardware Team Autonomy
3. Accurate Device Approval
As a lead designer, I proposed and implemented an innovative solution to address these challenges. Leveraging the power of AI/ML, I designed a web application that empowered the hardware team with the autonomy to test and approve devices for deployment.
Success Metrics
1. Time on Report and Approval
2. Hardware Team Autonomy
3. Accurate Device Approval
Impact
In just over a month, the hardware team conducted over 43 device tests, meticulously identifying 82 errors and approving only 2 devices. This rigorous testing, facilitated by the web app, not only elevated device quality but also introduced a heightened level of accountability into the approval process.
We streamlined the report review and approval timeline, slashing it from 7 days to an impressive 1 day. This empowered the hardware team to independently conduct tests, eliminating the dependence on data scientists, enabling AirQo to establish its presence in 8 countries across Africa.
In just over a month, the hardware team conducted over 43 device tests, meticulously identifying 82 errors and approving only 2 devices. This rigorous testing, facilitated by the web app, not only elevated device quality but also introduced a heightened level of accountability into the approval process.
We streamlined the report review and approval timeline, slashing it from 7 days to an impressive 1 day. This empowered the hardware team to independently conduct tests, eliminating the dependence on data scientists, enabling AirQo to establish its presence in 8 countries across Africa.
Context
What's AirQo and the Air Pollution Device?
AirQo is a Google-funded AI Research project — leveraging artificial intelligence to reduce air pollution throughout Africa. Our flagship product, the AirQo Monitor, undergoes meticulous manufacturing, testing, and approval processes before integration into the AirQo Network.
The device seamlessly streams data, which is meticulously calibrated by AI/ML data cleaning models, and subsequently disseminated for a myriad of impactful use cases.
AirQo is a Google-funded AI Research project — leveraging artificial intelligence to reduce air pollution throughout Africa. Our flagship product, the AirQo Monitor, undergoes meticulous manufacturing, testing, and approval processes before integration into the AirQo Network.
The device seamlessly streams data, which is meticulously calibrated by AI/ML data cleaning models, and subsequently disseminated for a myriad of impactful use cases.
Research and Strategy
Contextual inquiry
Conducting a contextual inquiry was integral to obtaining an in-depth comprehension of the Device journey, encompassing its progression from the manufacturing phase, through rigorous testing and approval procedures, to its ultimate deployment on the AirQo Network.
This informative session involved the active engagement of five participants, representing both end-users from the hardware and data science domains.
Conducting a contextual inquiry was integral to obtaining an in-depth comprehension of the Device journey, encompassing its progression from the manufacturing phase, through rigorous testing and approval procedures, to its ultimate deployment on the AirQo Network.
This informative session involved the active engagement of five participants, representing both end-users from the hardware and data science domains.
Structure and IA
Design Sprint
My objective for a design sprint was to develop a simple web app that will empower the hardware team with autonomy to test and approve devices while abstracting the complexity of ML processes in the background.
And the app aims to streamline the hardware approval process, ensuring timely and accurate device approval without the need for manual reports or the expertise of a data scientist.
My objective for a design sprint was to develop a simple web app that will empower the hardware team with autonomy to test and approve devices while abstracting the complexity of ML processes in the background.
And the app aims to streamline the hardware approval process, ensuring timely and accurate device approval without the need for manual reports or the expertise of a data scientist.
Rapid prototyping
Concept exploration and concept testing
During the design sprint, we delved into a variety of conceptual ideas, drawing from our prototypes and user flow explorations. At a critical juncture, I extended an invitation to our data scientist, seeking their invaluable expertise to evaluate the direction we were taking.
This collaborative effort aimed to elicit immediate reactions and insights, ensuring that our design aligns seamlessly with data-driven considerations, ultimately fostering a more robust and user-centric product.
During the design sprint, we delved into a variety of conceptual ideas, drawing from our prototypes and user flow explorations. At a critical juncture, I extended an invitation to our data scientist, seeking their invaluable expertise to evaluate the direction we were taking.
This collaborative effort aimed to elicit immediate reactions and insights, ensuring that our design aligns seamlessly with data-driven considerations, ultimately fostering a more robust and user-centric product.
Testing & Iteration: MVP + User Acceptance Testing
Deployed an MVP and Conducted User Acceptance Tests
We deployed an MVP to the hardware team, conducting a 14-day User Acceptance Test involving 7 devices with 3 participants.
Our primary objective was to assess the overall user experience, which entailed participants navigating the MVP, interacting with its features, and sharing feedback. We also aimed to validate specific use cases, where participants selected devices, designated test periods, and executed tests.
We deployed an MVP to the hardware team, conducting a 14-day User Acceptance Test involving 7 devices with 3 participants.
Our primary objective was to assess the overall user experience, which entailed participants navigating the MVP, interacting with its features, and sharing feedback. We also aimed to validate specific use cases, where participants selected devices, designated test periods, and executed tests.
Polish and Handoff: Post MVP Testing
Polishing the MVP
Building upon the insights gained from the User Acceptance Test, we embarked on an iterative journey aimed at enhancing usability and accessibility aspects.
Additionally, we seamlessly integrated the app into the latest AirQo Design System, ensuring a cohesive and visually appealing user interface. Through this process, we diligently streamlined the overall user experience, placing a strong emphasis on delivering a delightful interaction for our users.
Building upon the insights gained from the User Acceptance Test, we embarked on an iterative journey aimed at enhancing usability and accessibility aspects.
Additionally, we seamlessly integrated the app into the latest AirQo Design System, ensuring a cohesive and visually appealing user interface. Through this process, we diligently streamlined the overall user experience, placing a strong emphasis on delivering a delightful interaction for our users.
