Search

Apples can be Sustainable - 2020 Microsoft Hackathon

Learn about the projects we led and conducted at the Microsoft 2020 Hackathon

Every year Microsoft conducts a Hackathon where their employees are given the opportunity to take three days to conduct research and implement a team-based project of their choice. Employees have the ability to ask to join a project based on their personal interests and then the project creator has the opportunity to accept individuals based off of their credentials or ability to contribute to the project. During this Hackathon, innov8.ag was fortunate enough to be the creator of two different projects, "Apples can be Tasty, Crunchy, and Environmentally Sustainable!" and "Connect Farmers, Drones, & Farm Equipment with Private LTE and Azure Stack Edge." To learn more about these projects, please continue reading below...

Statistics on Hackathon Participants:


The Microsoft Hackathon focused on thousands of employee-nominated projects, with contributors from dozens of countries across the globe. innov8.ag was fortunate to participate in this event with two in-depth projects focused on apple orchards.

Apples Can Be Tasty, Crunchy, and Environmentally Sustainable!

Measuring and demonstrating sustainability has shown to be a significant challenge. As a Microsoft partner, innov8.ag is developing effective and efficient means to measure environmental impact, starting with utilizing Farmbeats to capture environmental sensor data at individual growers. We saw this opportunity of pulling together accredited data scientists, engineers, project managers, and software developers from Microsoft as the means to demonstrate this significant challenge. The primary goal of this project is to look at current, and past irrigation data in order to develop continuous projected needs on how much, how often, and how long to irrigate your crops. The team accomplished this goal by building a PowerBI-based irrigation insights dashboard that brings in farm-specific data from grower sensor & ERP systems, and aligning to KPIs that matter for different personas - grower vs. supplier vs regulators. The hack challenge was to build a architecture that supports both identifiable and de-identified data views, embraces the requirements of each persona, and succinctly summarizes environmental KPIs that are relevant to Microsoft's food-related customer base. In addition to creating a world-class sustainability dashboard, the goal was to build ML models that substantiate what inputs growers can change to affect environmental KPIs while maintaining the same or better outcome for crop yield (quality & quantity). Watch the below 3-minute video to see how we progressed as a team -




Our Hackathon Team:


Deanna Garcia- Principal Project Manager

Deanna was the principal project manager and Microsoft organizer, contact, and overall project manager for this group of individuals.







PowerBI Development:

Adrian Dobles Elizondo- Business Program Manager

Designed project plan for PBI branch, gathered customer insights and helped narrow down the KPIs, he also led and guided the project development in a strategic way.







Modhurima- Agriculture Economist, innov8.ag

Economic impact of data and data scientist- Modhurima focused on the economic KPIs, and how much chemicals we can save. Precision Ag helps reduce chemical usage, on average, by 4%. Less chemical usage will lower the cost of production and the impact of chemicals on the environment. Average labor hours saved is around 22% compared to traditional agriculture. Application of precision agriculture enables growers to retain soil moisture more efficiently than the traditional approach. Growers can respond according to the sensors and facilitate irrigation. The soil salinity level- a measure of nitrogen- can be adjusted better under precision agriculture as opposed to the traditional method.


Avinash Modi- Subsidiary PMM

Avinash Modi partnered with Modhurima Amin and Andres Dobles de Elizondo to analyze and quantify the Economic Impact of utilizing AI and ML technology to the apple farming industry in terms of improving crop yields with precision farming techniques. 




Paco Cruz- Sr. Data & Applied Scientist

Worked in PowerBI, he ultimately created queries to show soil moisture, temp, humidity, dew temp, soil temp, GDD Season (cumulative) all on one dashboard. With this consolidated view, growers can view many of the inputs that inform their weekly irrigation scheduling decisions.

  • Data manipulation and normalization to aggregate data from dissimilar devices

  • Feature engineering to generate budget line and irrigation volume

  • PowerBI model design, visualizations, and POC dashboard publication



Machine Learning and apple counting development:

Harmony Liu- Data Scientist at innov8.ag

Harmony Liu is one of the innov8.ag team members that assisted the Microsoft team during the process of the Hackathon. Harmony provided data support while working closely with both branches. She led the project on cluster analysis for soil zone identification, and worked closely with Peder on apple counting.




Peder Olsen- Principal Researcher, Azure Global

Led the the machine learning project. Proposed an application of neural network on fruit counting in the tree fruit industry. Peder created and used code to count annotated images of apples. which he then split the images into a testing data set to use a model that created image density maps.






Saptarshi Chaudhuri- Sr. Data & Applied Scientist

Soil Type Zoning: A machine learning based clustering of soil types based on different chemical and physical characteristics of the soil. The clustering model proves that soli characteristics are non-uniform and different zones in the soil have different concentration of chemical composites. This model will thus help plan for efficient fertilizer application and irrigation planning for the different soil zones based on the chemical content of that zone. Data aggregation on the PBI branch.

project: identify spatial varied zones in farm based on soil sample tests using clustering algorithm (one machine learning algorithm). 

Takeaways- Vet with soil scientists, drop Ca from analysis, remove Ca outlier, calls to action that soil is non-uniform, so uniform application of chemicals is not advisable, combine physical properties of the soil along with chemical composition

Able to identify two differences in the clusters based on pH levels and high calcium contents.


Shahab Morabi- Sr. Data & Applied Scientist

Shahab is a Data Scientist with the Business Application Group. He's passionate about solving sustainability challenges with modern technology. During the hackathon, Shahab was part of the effort to develop an AI model for automatically counting apples in orchard images. Put the data from the annotated images into the right format so that the models could be trained. He then used a GPU to allow the process to be quickened.



Feel free to share this post with a friend, family member, company, or anyone that you think could benefit from the information provided. Connect with us and please contact us through our website if you have any questions.

Follow us on Social Media!

Click on the icons below to be led to our social media pages!

Receive Our Newsletter!

Weekly Updates, Webinars, & more