AutoCode

Architecture Diagram
Problem Statement

Our Inspiration

COVID-19 has forced most of us to work from home for the last 3 months or so and now we realize that it may soon become the new normal. This transition has brought new challenges to our workdays. At the end of each day, many of us have felt burnt out and we have been looking for productivity-improving tools and techniques. All of us felt the need of a personal assistant, who could help us summarize the meeting points in a deck, who could create flow charts/architecture diagrams out of handcrafted notes and send it to different stakeholders, who could help in creating mock-up web pages, and so on. While such an assistant would always be an asset; in current times, it can prove to be a game-changer.

What it does?

We as developers, always have tools and skills to automate a portion of our tasks, but what if we could give a similar power in everyone's hands? Users can now take a picture of their meeting notes/ hand-drawn mock-ups and send it to AutoCode. AutoCode will read the image and extract relevant information from it. It then creates a PowerPoint deck or a web mockup for you and emails it to you.

AutoCode is building a modern productivity suite for workers. All of us would want people to focus on what matters instead of wasting their time on mundane tasks, such as formatting slides, aligning shapes, annotating images, reshaping training data, etc.

How we built it

We used a combination of Computer Vision, RPA and Python language to build the AutoCode solution. The solution can be divided into 3 modules: A2019 RPA code, Computer Vision (Google API) python script, and another python script that generates the PowerPoint/web page. The A2019 code lies at the heart of our solution. It orchestrates the entire process and decides the flow of the process. The computer vision (Google API) python script extracts the text and coordinates of the entities present in the input handwritten image and saves it in a form of structured excel table (xlsx). The PowerPoint creating python script uses this structured output and maps out the entities in a PowerPoint template, thereby, creating a deck.

Challenges we ran into

As a team, we all agree our biggest challenge was to work remotely and stay connected to make sure that we develop this solution efficiently and in time. Our other challenge was to find and use the appropriate technologies to get the desired results. We tried multiple Computer Vision APIs by providers such as Microsoft, AWS, Google, Abbyy, etc. to arrive at a decision of using Google Vision API. Also, creating the logic to convert structured data into PowerPoint and HTML codes was a challenging task. In addition to these challenges, we had many plans for our end product, all of which could not be possible within this time but we would surely want to work on them in the weeks to come.

What we learned

First, of all, we learned, working, and collaborating in this new normal. Probably, it was the first hackathon, that we participated in, remotely. It was difficult at times, but an enriching experience altogether. Secondly, we learned that the automation use cases are more obvious and evident than what we think them to be. At times, we are just stuck with point use cases that come from our client discussions and ignore the broader use cases that could have a much wider impact on society and business. Thirdly, RPA, along with other cutting-edge automation and digital transformation technologies, has great potential in improving business efficiency, enhancing employee productivity, and driving innovation across industries. Last but not the least, this hackathon has enabled us to think more innovatively and focus on the problem at hand; which many times gets difficult with the 'Business as usual' ecosystem we tend to work in.

What's next for AutoCode

Today, AutoCode is an innovative product and we have plans to scale it to become a generally available product for all. Initially, we plan to pilot the solution for a few companies and customize as needed to align with their templates, company logos, client logos, architecture diagrams, process flows. With time, we plan to understand the nuances of how we can scale this product to all the companies and more users, We will make required upgrades to the product and roll it out for the broader market. The product will be equally relevant in the post COVID world, as it is today, even if we assume an unlikely scenario of things to get back to pre-COVID normal.