An Update on Eetr and Squidgies

It's been three months since I embarked on my journey to building "something that matters," and I thought I'd share an update on the progress. This journey started, in my head, long before I quit my job. In 2020 I studied for a business certificate at UC Berkeley Extension, which opened my eyes to the possibility of doing something aligned with my values and could help those who matter to me. Then it was a matter of putting together an exit plan for my current job and getting involved with the activities of this new world.

It was shocking to see all the support I got from my friends, family, former coworkers, and acquaintances for something that seemed crazy. First, quitting a growing career as a Director in one of the greatest companies in Silicon Valley to develop a vague idea on how to help people reduce food waste and eat better.

In the beginning, it was an uphill battle; I had to remove the rust from my coding machinery and refresh that long-forgotten linear algebra, calculus, and statistics concepts. I had to learn what a neural network was, how it all tied with deep learning and everything I could about natural language processing. On top of this, I had to leverage new deployment models for apps that I knew only in theory, learn python and its relevant frameworks, and finally, refresh my web development knowledge, which was also ten years out of date. Not to mention building a useful app.

What have I achieved so far?

I've been working on two projects. First, building a machine-learning powered culinary platform, codenamed 'Eetr Platform,' which has many exciting features, and finally, it's starting to show a consistent end-to-end story. And second, I've been collaborating with my friend and former MuleSoft colleague to help him build "Squidgies," a machine-learning-powered language platform.

The Eetr Platform

Building this platform proved challenging because I'm trying to solve a problem that's not very well understood: Food waste and how it relates to people's health, farming practices, and shopping habits.

About two months ago, I got some help from my long-time friend Leo, an expert iPhone app developer, and we built an experience centered around ingredients. The app is a starting point where we can help our users do more with their ingredients. This approach represents a radical twist from what I initially intended, but now we're at the stage where we have an internal beta program, and we're very close to releasing the first use cases. If you want to participate in the beta, please let me know!

Bringing this app live significantly challenged my thinking about software architecture and performance. First, I had to figure out how to run state-of-the-art machine learning models on various platforms, including Google Cloud Run (serverless), keeping the costs at bay and good performance.

Squidgies

Squidgies presented a different challenge: I had to become a data scientist. While I haven't done much development work for this site aside from localizing it and maybe building a few tools, I had to read many research papers on Grammatical Error Correction, test and benchmark several models, and try to come up with a model that could correct grammar in 4 different languages: English, Spanish, French, and German.

I had to understand how deep models for translation and text-generation work and finally bring the GEC system to a state that delivers close to state-of-the-art results in all these languages.

Finally, I also brainstormed countless hours with Dan (Squidgies' founder) on what language learning meant and how we could deliver our users a platform that significantly improved their experience and supported their learning. If you're interested in trying it out, please follow this link.

Learnings and Thoughts of Where the Industry is Going

To close this update, I would like to share my thoughts as someone who has observed the software industry for over a decade.

In the same way that integration tools blew my mind a decade ago, and I could see more companies building their businesses by combining many existing solutions, now I think it's time for machine learning to shine. Machine learning techniques can help consumers and businesses achieve things that seem impossible so far. These technologies were available only to some elite companies during the 2010s, and now they're becoming a commodity that anyone can take advantage of. So there's no surprise when you see companies that process large quantities of data flourish.

As an example, in my app, I leverage many ML techniques to change how users search for data:

• Voice recognition through Siri allows users to search reliably without typing.
• Models like BERT allow my app to extract meaning from text strings and translate that meaning into a complicated search query for ingredients.
• Semantic search allows my app to find ingredients by meaning and not as a plain sequence of characters.
• Decoder-style transformers leverage massive internet data to generate creative suggestions and reliable recipes.

With all of this, I'm excited to see what the future and AI bring to the industry.