Recipe for Success — Start Cooking!
It’s love. Always add some. Enjoy!
Building a startup is very easy. You can do it too. Let’s do it.
- Define a problem
- Get some data
- Write some code
- Run that code
- Use the result!
Simple. Isn’t it. I knew you would agree with me.
So why this article is so long? Well, we completed building it, now have to run it. Right? Let’s start by rolling the carpet in your new office.
Define a problem
This is the part that you, your senior management and product managers do. Yes, those long meetings. The same ones where we use spreadsheets and take a lot of notes. Everything is important, but here is all the power to change everything else. It involves multiple processes but few notable ones are surveys, feedback, domain research, yada yada. The critical tip here is to manage information flow. Listen to everything possible, filter out, generalize and conceptualize.
Get some data
Well this is abundant, right? Not so much. It’s the same case as we talk about water. 71% of earth is water, yet people are dying of thirst. This is where we try to determine which water is potable. Which one is treatable and treat it. Desalination is also possible, but costly. Crawlers do that every day. Big corps do this on their own data every day. A very general way to approach it is that data is never clean enough. One tip here is to try to have a feedback loop running. This way you can tell data source that if they used glass based water pipes, you will never have any issues. They never listen. Tell me about it!
Write some code
You might have heard of this term: Machine Learning. This is the buzz these days. Everyone’s doing it, yet we don’t have personal droids as good as R2D2. He is the best. Alright, don’t start a war, BB-8 is good too. Happy? Well, back to the our galaxy. There have been many developments in this field and recently it’s been slowly made easier to approach. It’s a dragon, beware. This is where those PhD guys who spend days circling research papers and circling some more on whiteboards come in. Don’t worry. They are just trying to do one thing. Remember the problem that your manager defined, and data guys gave data for, guess what, we did research for 30 days and prototyped these 5 models (essentially code for whatever your manager said in english). Oh, did we mention, we will need more data.
Run the code
Well this is the easy part, right?
source awesome_model.awesome_language Right?? Well, as it turns out. Not so much. You need lot of raw computation power, environment setup, delivery mechanisms, data pipelines, failover, replication, redundancy, versioning and what not. Well, have you seen some guys in your company who are always frenzy and always looking at some graphs which look dangerous and always have a terminal open. What do they even do in that terminal? Well, they are taking care of the command we wrote above earlier keeps working.
Use the result
Finally, time to show the magic trick. This is where the product gets built. You wanna use the AB test ran last month to improve your conversion rate. Or maybe you wanna use that natural language processing model that is ready now. This is where everything becomes as easy as saying “Ok Google, Tell me a love story”. This is what majority of current populus does, don’t ask me what they do. Everyone’s got to use promises. Why won’t they, sigh!
Well, here we are. All done! Smiling like an idiot at the IPO. And then the dream is over. With sad realization that you will have to build teams of hundreds of engineers to achieve just this simple dream of yours. Sigh!
Let’s use your “5 minutes of extra sleep” time to make that dream easier, if not very possible.
- Problem: Can’t help here bud, you are the one with million (rather billion) dollar idea. Go get it! No? Go through these sometime.
- Data: Go big!
- Code: Go easy.
- Run: Be resilient.
- Use: Again all yours, but I can help. Why not a chat bot? No? Sure?
Now that you have all the links open in a new tab let me just make it simple for you. Google Cloud Platform uses NoOps approach with almost all of their products. With that you reduce the size (hence worries) of your team. You can start yourself!