Product Development @ Acerta
Wearing Many Hats
As with any small startup, the experience of working at Acerta has been one of many opportunities and tons of learning. During my 2+ year stint at the company, I have officially filled the role of front-end software engineer but unofficially been involved in product design, product planning & vision, software architecture, devops, solutions work in the way of demos for pitches, mentoship of interns, and more. Because it's a lot, I will break down my work into digestible chunks and talk about the various things I've learned on my journey at this company.
Product Planning, Vision, and Design
What I consider my most significant contriution to the company, I have a relentless passion for driving the company's product forward. I have constantly evaluated the existing designs and, where necessary, extended or completely overhauled them. In a number of scenarios, my decisions have changed the way the entire company approaches a problem.
Development of a "Module-based" Platform
One of the challenges the company faced with its main product, LinePulse, as it grew was that this general-purpose tool for reporting on machine-learning-driven insights on the assembly line meant different things to different customers. For example, one type of customer saw the value in LinePulse for its "Anomaly Detection" capabilities; looking at single units as they pass by an operation and flagging problematic ones. However for other customers, their problems aren't just at one operation, they're spread all over the line so they prefered LinePulse's "Root Cause Analysis" capabilities.
Though in retrospect, it's painfully obvious that LinePulse's feature set needed to be split into modules, this was not so clear at the time. Our entire infrastructure from the data science models, to the database, through the back-end and to the front-end was set up on the assumption of a general use-case that simple didn't exist in practice.
At the beginning of 2020, I proposed a module-driven based design for the platform along with an accompanying UI redesign that had immediate buy-in from the company. Throughout Q1 and Q2 of 2020, we worked extremely hard to integrate all of our existing LinePulse functionality together into one platform. The payoff? I no longer have to completely redesign the product to match new use-cases; we either use an existing module or build a new one!
On a side note, I'm very proud of the fact that this module-based approach has permeated Acerta lingo. Everyone at the company knows what we mean when we say "RCA" (Root Cause Analysis) and "AD" (Anomaly Detection). Marketing materials now reflect our module-based approach and even the database has module-specific tables!
Rebuilding the RCA Module
Shortly after completing all of the core modules of the new LinePulse platform, a sales engineer and I came to the conclusion that the RCA module was just not cutting it in its current form. We had received lackluster feedback from our existing and potential customers regarding the insight they felt they could derive from the module; to make a long story short, they felt it wasn't enough.
Compared to all the work we had just done migrating LinePulse to a platform-based approach, the RCA module we concluded we needed was a whole other ball game. We would give our clients the ability to explore their own data and draw their own insights from the results generated by our clustering and classifier algorithms. The new UI would need to have highly interactive plots, the ability to visualize the results of our analysis in different domains (aggregates and sequential), and compare data between various features and clusters.
Encouraged by a mentor and supporter at work, to figure out what exactly to build and to get buy in from the whole company, we introduced the idea to people gradually with a short sales pitch and then collected their feedback and input. Over the course of just over a month, I set up meeting twice a week with someone new from the company to do just this. In between sessions, I would update our mocks with any insights I had learned. Finally, once the design had become clear and we had received buy-in from many important parties, my design was positively received by the entire company and we began planning the implementation.
Throughout this process, I realized the value of knowing how to sell my ideas, especially ones that are more difficult and/or controversial. Due to the fantastic guidance of my mentor, I gained a toolkit for doing just that.