I open sourced my startup | Hacker News Reader

I open sourced my startup

42 pointsMikeChristensen16 comments

So I've spent the last 3 years attempting to teach a computer how to understand recipes. What they taste like, ingredients you'd need to make them. If they have animal products or are gluten free. If they're a dessert. This work involved a lot of natural language processing, Beyesian filtering, etc.

With this, I built a very powerful recipe search engine. It allows you to find recipes using some pretty interesting queries. For example, if the user wants to make 5 recipes that will efficiently use a head of lettuce, 12 bananas, a pound of salmon, and 6 eggs, it can find the most efficient set of recipes that will use as much of those ingredients as possible, while requiring the user to have to buy as few new ingredients as possible. It's also able to normalize the ingredients across all recipes to build really accurate shopping lists. For example, if one recipe calls for 6oz of cheddar cheese and another calls for 1/4 cup shredded cheddar cheese, it knows the weight of 1/4 cup of shredded cheese and can combine that with 6oz. So, pretty interesting technology.

Unfortunately, after several iterations, this completely failed to gain any traction as a consumer product. However, I got dozens of emails from other entrepreneurs who were trying to build similar products and were very interested in the technological side of things. I had a lot of asks if I could share the code, or turn it into a web service.

For this reason, I have decided to re-factor my code as a free, open source project. I'm hoping to find some other developers who could build on what I've done to create the next big recipe startup. The code is all written in C#, but could almost definitely be run under Mono.

It's up at: https://github.com/KitchenPC/core

If you'd like to help contribute, by all means! I really hope it can help someone out!

Mike, its pretty sad that you had to close this down. I work at Cucumbertown, http://cucumbertown.com/ that’s pretty close to what you do but not exact. We don’t enter the search vertical but focus just on building this as a replacement for food bloggers. Jef Miller (Punchfork acquisition to Pinterest) wrote something very similar for Pinterest http://techcrunch.com/2014/01/21/courtesy-of-punchfork-acqui... though not exact.

From my learning: Build a solid BD team that can focus on taking your core value proposition to the masses. As a founder create the story. The market is flooded with recipe related products that exist since 1995. Unrelated but I wrote the first V of Cucumbertown in C#. Then had to port to Python.

I hope you don’t lose the enthusiasm to continue keeping this open sourced.

The big problem with recipe startups is that it's incredibly hard to make money.

Most startups either go the route of affiliate sales (incredibly hard; supermarkets make tiny margins as it is) or brand promotion (which is better but you still need a huge audience and dedicated sales team to make it work).

I can't think of any recipe focused startup that's managed to raise money from top-tier VCs (although some have used it as lead gen method for other businesses).

I work at Cucumbertown: http://www.cucumbertown.com/ and we’ve managed to raise from venture funds. We are not just recipes though. It’s more about the cooks: http://jane.cucumbertown.com/. There are a lot interesting models outside of ad & affiliate sales.

Kinda related, but I know there's a pretty decent-sized market for startups/companies/groceries that need barcode data for the entire universe of foods for their apps (Think MyFitnessPal). I know of someone making 6 figures a month selling this data (he makes 5 figures per sale on average), and he's probably one of the few reliable data suppliers.

So for anyone looking to get into this type of market, that's an idea for you.

Clickable: https://github.com/KitchenPC/core

Could I ask why you didn't turn it into an API and pivot into an API as a service?

Yes - very straight forward answer to that: I did a lot of interviews with potential customers, and it was a deal breaker for virtually all of them. Basically, no one was foolish enough to build a startup that depends on a service which may or may not stay up and running, or go out of business and disappear completely.

Thank you for sharing. I've been thinking about building something similar for personal use just to make doing the groceries more efficient.

I'm also starting to learn machine learning so this should even make this mofe interesting.

This is awesome. Always a fan of people open sourcing cool technology like this!

Is there any chance to open source collected recipe data as well?

This is a tough question, which I don't have an answer to without throwing out "IANAL" type acronyms. The recipes I've collected were crawled from several different major recipe websites. I think my website database has somewhere around 50,000. Not only would a sample database with that many recipes be huge, it also made me uncomfortable to include content scraped from AllRecipes, Food.com, Epicurious, etc in an MIT licensed open source project. Do I have the rights to even do this? Are there copyright issues involved? Who knows. The best approach, at least for now, was to include a couple dozen hand-picked sample recipes (which I got from friends and family) as a starting point. After all, I'm open sourcing technology I built, not recipe content I did not create.

I totally understood your point. Besides a project can have several aspects. Somebody can have interest over your algorithms, others can have interest over your processed data. On the other hand, I don't think recipes are copyrighted. Actually I am not an expert on this matter but most of time, also you might realize that famous sites use copy-pasted or slightly changed materials. Inventing a recipe usually is not a case. If someone can enlighten us on this subject, I will be really appreciated.

Thank you for granting the world the spoils of your hard work.

No problem!

is KPCData.xml the whole recipe database?

It's basically sample data. I hand picked about 30 recipes from my database, at least one from each category. The sample shopping lists, menus, and favorites also use those recipes. Not enough to build a massive recipe website with, but enough to get you started.

Reply on news.ycombinator.com