Digital Sense

BACKGROUND

Introduce your business and what you do there.

I'm the CEO of Software Testing Bureau, a software testing company. We offer a software testing automation tool that's easy to use and accessible to companies of any size and sector. Our tool helps improve productivity and quality in development and digital transformation processes.

OPPORTUNITY / CHALLENGE

What challenge were you trying to address with Digital Sense?

We needed to implement artificial vision into our tool so that our software robots could deal with changes in the applications being automated.

SOLUTION

What was the scope of their involvement?

Digital Sense implemented and trained a machine learning model that allowed us to recognize elements on the screens and use that information to control a robot when the application changed. Later, that same characteristic was coupled with a large language model (LLM) that allowed us to build robots from a user description of the process to automate.

What is the team composition?

We worked with three teammates from Digital Sense.

How did you come to work with Digital Sense?

We found Digital Sense through a referral. We chose them for their expertise in artificial intelligence and experience with artificial vision.

What is the status of this engagement?

We worked together from September 2019–August 2021.

RESULTS & FEEDBACK

What evidence can you share that demonstrates the impact of the engagement?

We tested every delivery, so we had a clear understanding of whether we were going in the right direction. We used more than 900 screens to train the ML model and around 250 to test the model's capacity. As a result, our tool now has the ability to see and understand an application screen.

How did Digital Sense perform from a project management standpoint?

Considering the project's difficulties, Digital Sense delivered on time. We communicated through meetings.

What did you find most impressive about them?

Digital Sense's knowledge and expertise were impressive.

Are there any areas they could improve?

I can't think of anything for Digital Sense to improve.

BACKGROUND

Please describe your company and your position there.

I am the general coordinator of the Agricultural Development and Climate Change project, between the Ministry of Livestock, Agriculture and Fisheries (MGAP) and the World Bank. The responsible institution is the Directorate of Natural Resources of the MGAP

OPPORTUNITY / CHALLENGE

For what projects/services did your company hire Digital Sense, and what were your goals?

Agricultural development and climate change project, within the framework of this project, the DSense service was contracted to develop a web computer system that made it possible to make the soil conservation policy defined by the MGAP viable.

SOLUTION

How did you select this vendor and what were the deciding factors?

An international public call was made and the company was selected for its best solution proposal to the MGAP need.

Describe the scope of work in detail, including the project steps, key deliverables, and technologies used.

1.  Technical approach and methodology 
   1.  Main characteristics of the proposed system The proposed solution tries to take advantage of the integration of the three modules (Entry, Verification and Inspection), to provide a tool that, in addition to correctly receiving and evaluating the data, provide help (destined to the Responsible Technician - TR), which makes the system friendly and collaborates in the correct assignment of the main parameters of the USLE-RUSLE equation on which the presentation of the PUMRS is based. This objective is, in our opinion, key to maintaining the massive adoption and minimization of claims in the first years of implementation.

The main characteristics of the proposed system are summarized below:

• Full Web System. All the data handled will be in standard formats (SHAPEFILE, KML, GEOTIFF, etc.) to guarantee interoperability with other systems, particularly ARCGis, the system used by RENARE. 
• Extensive use of the information available by RENARE for the entry and validation of the use plans presented by means of the automatic calculation of the main factors that make up the USLE-RUSLE equation. 
• Workflow-based system that allows a correct separation between the interactions between the different actors and the use plans presented. 
• Control of PUMRS through the use of satellite images and specialized detection algorithms, machine learning and pattern recognition. 
• System based on MODIS and LANDSAT satellite images, available free of charge.
• The system will also allow working with other satellite images in case access to them is obtained. The proposed system offers automatic solutions that facilitate each of the stages (Entry, Verification, Control) that were previously described. In addition, many of the requested components will be totally or partially based on dSense Agro, a product developed by Digital Sense Technologies (see background), with several years of testing and use, which drastically reduces the risks associated with the development of the themselves.

The components of the proposed system, which are based on dSense Agro, are detailed below:

1. Creation, loading and modification of the polygons that represent the different units of analysis in a GIS system (see background Applications for the elaboration of Plans Responsible Use of Soils). These components will be used in the information loading module.
2. Automatic calculation of the LS factor (see background Applications for the preparation of Plans for Responsible Use of Soils). This functionality will be included in the usage plan entry module for the USLE-RUSLE equation.
3. Settings for precision agronomy (see background Agricultural and forestry settings). Digital Sense Technologies' experience in the analysis of satellite images to identify productive environments will be used for both verification and inspection.
4. Estimation of yield (see precedent Estimation of pre-harvest yield). For this component, Digital Sense Technologies used MODIS and LANDSAT images to extract information on critical moments in crops. Something similar is proposed to monitor the use plans submitted, see Appendix: Proof of Concept Monitoring.

How many people from the vendor's team worked with you, and what were their positions?

4; Architect, Functional analyst, Programmer, Specialist in Machine learning and image processing

RESULTS & FEEDBACK

Can you share any measurable outcomes of the project or general feedback about the deliverables?

The system is working and is used by 96% of the target population.

Describe their project management style, including communication tools and timeliness.

Work was done on a hybrid of finished product and sprints. In addition to the usual communication channels, the Mantis bug tracker was used to manage the team's tasks. Their deviations in times were normal.

What did you find most impressive or unique about this company?

The ability to understand the specific needs of the MGAP and its solution proposal

Are there any areas for improvement or something they could have done differently?

Runtimes need to be tuned more