AUtomatic SPeech REcognition (AUSPRE)

(Pronounced Osprey, like the bird)

Source: Ivey Entrepreneurship Report 2020

Introduction

Fail fast and fail often is the idiom of startups. I never thought that it would apply to me and my startup, but here we are.

AUSPRE started out as a 3-man operation back in late 2019. We had a businessman with a vision, a technical lead with the experience, and an academic with the idea. Guess which one I was? I definitely was not a businessperson nor was I seasoned enough to call myself a tech lead. What I did have was the ability to take the algebraic jargon from NLP papers and implement the formulae into models that we could train and run inference on.

Like any other startup, we began with humble beginnings; we worked out of a shared office space one of the co-founders had access to, and had no initial funding other than the sweat equity we dedicated into the company. The first few months were a whirlwind; by following industry practices of agile software development, we quickly setup a CI/CD platform on GitLab that auto-deployed to staging servers hosted on DigitalOcean. Primarily we leveraged Rancher for our K8s infrastructure and Composer for modular dependency management.

Our stack was simple- PHP for the backend supported with Python for the AI inference components. A MariaDB database for storing transactions and edit records, as well as a NodeJS frontend. Our proof of concept went from a MVP to a fully featured end to end platform including stripe integrations that allowed customers to easily onboard and get started. It was around this time that we started considering avenues of funding in order to help bootstrap and accelerate the growth of our business. As the majority of us were based in London ON, it was only natural that we applied for the Western Accelerator program in 2019. I applied initially and was rejected; however after a second round of consideration, we were ultimately admitted into the program.

The program was awesome - the only requirement was that I had to spend 4 months working in the accelerator full time. In return, the accelerator would provide us with resources such as equipment, marketing tools and subject matter experts to help take our startup to the next level. During my tenure in the Western office, I met some of the most remarkable individuals- people with a passion for healthcare, logistics and commerce both B2B and B2C. In this period, we obtained contracts with educational institutions and onboarded multiple paying users. Business was great - so where did things go wrong?

As you probably have guessed by now, 2020 brought about unprecedented times- including a global pandemic. Along with the pressures of a locked down economy and reduction in spend from businesses, we were wrestling with problems of our own. It was difficult to source clean and reliable datasets to help augment our models. Speech to text is difficult if not impossible without large amounts of data- and it is very difficult to compete with companies such as Google with their endless catalogue of data stemming from platforms such as Youtube. Despite our models being transformer based, training times would often stretch into the hundreds of hours making it difficult to scale. We incurred significant technical debt by trying to modularize our code packages to the point where the limiting factor in our productivity was waiting on package builds. It's true that these challenges were difficult, but not insurmountable. The biggest factor in our failure to scale simply stemmed from creative differences on where our product was supposed to grow and develop.

While no founder wants to see their product fail, there comes a time when it's better to admit defeat and move on, lest you fall into the sunk cost fallacy. AUSPRE was a great venture into entrepreneurship. I learned an incredible amount from my colleagues at the Western Accelerator. I cherished the good times I shared with my co-founders, and struggled together with them through the hard times. Ultimately, we decided that it was best to take our learned experiences and move on from the company.

My fellow entrepreneurs and I at the Western Accelerator.

"It is hard to fail, but it is worse never to have tried to succeed." - Theodore Roosevelt