Darwin Reviews

I provide product management and SME services to oil companies as a consulting service. My company has partnered with SparkCognition to bundle its products into a package of services that I provide to my customers. For the most part, when I'm working with SparkCognition, and Darwin in particular, I'm working with it on behalf of one of my customers.

We do different engagements. We've done PoC projects with customers with versions 1.4 and onward.

The biggest use case we've seen is for automatic classification of data streaming in from oil and gas operations, whether exploration or production. We see the customers using it to quickly and intelligently classify the data. Traditionally, the way that would be done is through a very complicated branching code which is difficult to troubleshoot, or by having it manually done with SMEs or people in the office who know how to interpret the data and then classify it, for analytics.

The customers have looked at using machine learning for that, but they run into challenges — and this is really what Darwin is all about. Typically there is an SME who can look at the data and properly classify it or identify problems, but taking what he knows and what he does instinctively and communicating it to a data scientist who could build a model for that is a very difficult process. Additionally, data scientists are in very high demand, so they're expensive. 

SMEs can look at data and quickly make interpretations. They've probably been looking at the data for 10 or 15 years. So it's not a matter of just, "Oh, we can plunk this SME beside a data scientist and in a couple of months they can turn out a model that does this." First, SMEs don't have time to be pulled out of their normal workload to educate the data scientists. And second, even if they do that you end up with something very rigid 

With Darwin, customers can empower the SMEs to build the models themselves without having to go through the process of educating the data scientists, who may leave next week for a better paying job.

Most of the projects that we've done, PoCs, are typically done in the cloud, for ease of use. Because we work in the oil and gas space, public cloud is the preferred option in the U.S., with the simplified administration and a little bit lower cost. Overseas, the customers we've talked to have noted there are laws and restrictions that require their stuff to be on-premise. We've talked to potential customers about it, but we haven't actually done an on-premise project so far.

The automated AI model-building reduces the time that projects take. Before I started working with SparkCognition, I worked on several projects where it took months, and in some cases years — complex problems — for data scientists to even pick a machine-learning model to use. They might settle on a methodology such as random forest after quite a bit of analysis. Whenever a model is completed, it is a powerful and unique solution that can't be done with traditional programming, but it's almost impossible to tune in the field. Additionally, if you're talking oil and gas, some of the sites where you need to run these, especially on the edge, are very remote. It doesn't respond the way you want. So then you have to take it back to a data scientist and have it tuned. 

Darwin lets you rerun the process with new data, with more data, with different tuning parameters. You don't have any of that back-and-forth with a physical person.

The solution has created the opportunity for machine learning to be practically implemented in places where it couldn't be implemented before. The current way that machine learning problems are implemented is with data scientists, usually as IT initiatives or R&D initiatives. Often, a company will say, "Okay, we're going to do machine learning." They have a big initiative, then hire some very expensive data scientists and they create a model that may be 10 or 15 percent better than what's out there. The challenge is that the model exists in MATLAB or Python but it's not integrated into the business systems like an ERP; or it's not integrated into their industrial control system. It ends up being a really cool PoC and it never turns into something that practically affects the business.

Darwin has opened up the places where you can do that.

The potential we see with Darwin, with the REST API and with the easy-to-approach interface, is that we can empower these SMEs to build things and interface them from day one with the existing control systems and other systems the business is using. So they're not stepping out of their traditional workflows to use machine learning. It's integrated.

As far as building models goes, versus hiring people with ML experience it is significantly faster and exponentially cheaper. You can build models, once you've done some initial training in Darwin, that would take a data scientist, with an SME, two or three months to build. With Darwin the SME can do that in a few days or less. For a lot of applications, especially in oil and gas, the savings are huge as far as practical applications of machine learning go, versus the tradition of using data scientists to build them one by one.

For our customers we have primarily looked at use cases around automatic event detection. We hadn't even tried to do that with data scientists because it just wasn't practical with the timeline and because the costs were too high. And using traditional software methods to try to solve that problem, the estimate that the customer had was that it would have taken three to four months of software development. We were able to build a model that provided effectively the same results within a week. A lot of that was just figuring out the data and data quality issues. The actual model building in Darwin took a few hours.

We have been using it for our risk management portfolio. We are a lending institution. We give credit to small and medium enterprises. We've been using it mainly for client segmentation and the probability of delinquency in the loans that we get. 

I am using the latest version.

Due to the predictions that we have been able to do because of the use of Darwin, we have decreased our delinquency index. We were almost at nine percent by July. After using Darwin, we have reduced the delinquency index to five percent. Darwin is important for this metric.

We calculate the potential loss of clients. When some of our clients are going to go with competitors, they will ask for money with another bank or institution. In July, we had this number at 19 percent of our clients. We saw that they were likely to go with some of our competitors. Now, after using Darwin to narrow and segment more of our clients, we have been contacting them in a more specific way and targeting them more specifically. So, our clients loss index has gone down to 10 percent. This is also important.

I like the solution’s automatic assessment of the quality of data sets for uncovering issues like missing data points, low statistical variance, or incorrect data types because of Darwin's process capacity of filling in the data for us or even giving it double or triple checks. However, I still believe that the data cleaning process should be cleaner and better before it arrives to Darwin. We still have a lot of processes to do before we put the data sets on Darwin.

We have a process which gives us the credit score for all the credit companies in Mexico. We have to clean the data first through Darwin, so that we can use it in our modeling. Darwin has been very helpful on cleaning this data with simple queries.

We use it more for production of models. We have not had any trouble whatsoever with the data repository connection. It's consuming the data sets correctly, which is good.

It does affect the decision-making of our risk committee using the new models. They adjust the data accordingly, then this committee reviews everything each week so we can be fast enough to compensate for any new approach or findings that we have through Darwin. It is constantly optimizing the model, so once a week has worked for us.

Darwin allows us to make better decisions and more quickly.

I believe that Darwin is helping us to evaluate all our processes in a much faster way. That is why it affects our productivity for the good. It solves a lot of the time consumption. This really makes a difference for us: the time consumption. The processes, evaluation of the processes, and results may be the same, but the time it takes us to process it through Darwin rather than on traditional tools or other tools is so radically different. It is changing the way we make decisions because we can make decisions faster, and maybe, that's the most valuable thing about Darwin.

We use it for analyzing data and creating models. We extract information from the database and then see if Darwin can share information with us about what would be nice components for the model. Then we use Darwin to make a model. We clean the data and pass it through to Darwin and Darwin generates a best model.

From Darwin, we get parameters, important features, and predictions. We don't have the entire Darwin solution. We just have the core. We are taking the information about the parameters of the model and then we generate the model again with our own tools. Darwin doesn't give us the actual model to use, just the parameters.

We work with Darwin through a webpage and create models there and do linking analysis of the data. We are also working in the SDK version. We connect with the cloud, through the console.

When you start to use Darwin, you realize how the parameters improve the model. Darwin uses an algorithm that creates 10 or 15 generations and gives you a really accurate model; a more accurate model than we used to make. When we have a clean dataset, within two to three hours we have a really nice model, one that is better than we could generate in a week. A couple of hours with Darwin is like a week of work for us. That's where it provides the most value: in saving time for us. 

We are using it in two ways. One is by analyzing our current clients to create more business by deciding if we can offer them new products or if there is a risk of their leaving us or stopping use of our credit lines.

The second side is to prevent the risk of default. Our credit clients, because of the economic situation or internal decisions of the company, can go into default and stop paying their credit lines. We use it to prevent that risk. If we see a deterioration in a client, we can decide to stop lending money to the client and prevent risk in that way. 

So on the one side it's to create or attract more clients by identifying certain trends or certain characteristics and offering them more products. And on the other side, it's to prevent the risk of credit default.

Both of our use cases are valuable. On one side it creates more business, but on the other side, it prevents impact on our company's equity. We have seen positive results on both sides, but especially on the risk prevention. Still, we have seen positive results in the generation of more information for our commercial team to attract clients as well. We have managed to reduce the percentage of high-risk clients from 8.9 percent to 5 percent. That's very significant in the overall quality of our clients. That's very good news for everyone, especially our investors.

Darwin is also helping us to make models faster and increasing the efficiency of the team. Instead of trying and trying to prove various models, we get a model fast. It's making data science faster but we would like it to be even faster still.

It has helped our team to convert data into knowledge and to more deeply analyze the results.

The primary use case is to predict the default on payments by clients.

We have a lot of data sets with different kinds of mistakes. We're using Darwin to help us to fill in the blank spaces. It helps us to solve that problem and clean the data when we have blank spaces in it.

We are increasing our number of operational algorithms. We are developing algorithms to predict data for the next month, including forecasting our income using macroeconomic variables. About 80 percent of our models are operational.

We are able to develop models faster. We are a small team; there are three of us involved in business intelligence. We develop models faster because, instead taking three weeks to test a lot of algorithms to select the best one, we just upload information to Darwin and let the software help us define which is the better algorithm for that data set. We developed a model for a customer and we are on track to have a new version of that model in two months. We can do that faster because of Darwin.

Darwin enables our company to tackle more complex problems by making data science more approachable and operational. While we need people who understand what's happening in the business, the solution helps us to use our time better. We need to be more creative to have more variables that we can use for making models. We are working on creating bigger and better data sets, to increase the quality of our data sets, and on the automation of the model and the result of the model.

It has increased both efficiency and productivity. It's helping us to improve our data sets and add new variables, instead of proving and testing a model for multiple weeks. Our productivity is increasing because we are making better decisions. We are making a change in the company, becoming a data-driven company and really using our data better. For example, with our recent development for the sales team, there has been a 20 to 25 percent improvement in efficiency because we developed a solution that helps the sales team to identify when a client has enough money.

Our main goal is to transform data into knowledge and Darwin is definitely helping us to do that faster. An example is the issue of loan defaults. We have a lot of data on loans that we have already made to clients and, on some of those loans, the client defaulted in payment. We took that data and converted it into a model that helps us to predict which clients might default. We have started to cancel those loans and recover the money. Without that data, we would have a higher rate of payment default.

I have been working on data analytics using Darwin. I have been working more on the data generation part. There were some problems where they wanted us to generate some synthetic data, and I was working on that part. 

As for the usage of Darwin, somebody else does that, but I also am getting familiar it. 

We were using the last version before 2.0 was released.

We just load the data, then analyze the data. We just check if some of the data is missing or if there are some outliers in the data. We can analyze the data to see what are the relations between the columns of data.

We create the data and models with Darwin. We use those models and data in products. It's a great help.

It saves some time. Even people who are not fully technical can use it with a little guidance or something. It increases productivity.

Initially, I worked on creating a pipeline. I did this once. It has automatic AI building in its API pipeline.

The PoC we did was for the oil and gas field mostly, as well as the aerospace field, to optimize supply chains. We wanted to see what level of information we could gather from using this tool and how it would help us. We were looking to become a reseller for Darwin and to provide services through it to our clients.

We wanted to pitch it to our clients, but our PoC indicated it was not feasible.

It's a good tool for a beginner business analyst. I don't think it's good for advanced analytics.

When I created a model, versus when Darwin created a model, my model seemed to perform better, simply because I have more knowledge about the actual business policies and problems that we have. Darwin was just giving me a beginner, data science-type assessment.

It might have improved our ability to tackle more complex problems by making data science more approachable and operational but not by much. It improved that by about 1 or 2 percent.

It didn't increase efficiency or productivity on our team because we had to spend more time trying to figure out how it worked rather than actually working with it.

It did help us to convert data into knowledge. It did provide a little bit of value.

I was trying to see if Darwin was going to be useful for the company and if it was useful for the project that I was working on. I was working with it, testing it, seeing how it worked, seeing how accessible it was, and if it would be something that would be viable for us to use.

We were hoping to use it on a machine-learning project, to categorize words based on their likeness to each other. I had to find a way to translate that, and encode it, into something that Darwin could actually read.

Darwin is really useful for people who don't necessarily do a lot of data science. A lot of what you're doing in Darwin, while it's a lot more efficient, you could do yourself if you had the knowledge and the time. It's definitely more efficient and it's also useful if you are someone who does not really understand how data science works, but you still want to implement some machine learning.

It gave me ideas for how I would want to implement something on my own. I wasn't going to use Darwin as my long-term solution for what I was going to do, but it gave me the ability to tweak it in different ways, and to do that a lot faster than if I was going to code it myself on Azure Notebook or AWS. I found that really useful. It made my workflow a lot more efficient.

My project, perhaps, was not the best fit for Darwin. It did not necessarily help me find the answer for that project, but it did allow me to work with different ideas of how I would want to set up the project and pursue those answers. So while it didn't hand me the answers, it gave me the sandbox to find that more correct workflow.

It saved time in two ways. One was that it made it more efficient to tweak some options and then run it. I didn't have to put in all the code myself or rewrite the code. And it also made it a lot easier to have multiple machine-learning processes run at once. Darwin has it neatly all in one place, which makes that a lot easier to use and, again, more efficient.