What we do
We at BIGmama Technology have profitably executed projects involving machine learning for the last six years. We have specialized in delivering bespoke, turn-key applications for large organizations in different sectors such as energy, telecommunications, banking, employment, and rail transportation to solve problems faced in diverse functions from sales and marketing, to rail operations and maintenance, to employment departments.
Why we're building this
We're building iko to enable us to Get Data Products Right: load and manage data; train, validate, deploy, invoke, and monitor models. Plug them into a modular application. We want shorter time-to-value for our clients. We want self-service for our data scientists and machine learning practitioners. We want near real-time collaboration. We want distributed. We want asynchronous. We want that all for yesterday.
We wanted to take our clients from puzzled look to wide smile as fast and efficiently as possible.
Challenges we faced before
Developing custom solutions for enterprise from initial meetings, to problem space exploration, to problem statement, to custom data sources connectors, to model training, to the web application that leverages the models that the domain experts will rely on, to the mobile application to visualize data collected from distributed, unattended devices, to the code running on these devices, to custom built hardware, to infrastructure specification, to model and application deployment and management obviously leads to problems one individual working alone on toy projects during the week-end does not necessarily face. That time is to be added to the time required to understand the problem and dive deep into a domain. We have done that with a relatively small team, which means that team members had to be able to move across these steps. Furthermore, it complicates hiring and team member retention as working on several simultaneous projects is polarizing: thrilling for some, excruciating for others. We only could tackle a few projects per year, too. Keeping different application code bases in our heads was taxing, and managing and supporting instances of these custom applications was a challenge.
Only very large organizations could afford our services, which meant that the sales process was slow, even though most of our clients were repeat clients. It was also difficult to amortize our engineering efforts in the beginning, as one project was different from the other. That is, until we have done enough projects in enough sectors that some patterns have emerged.