New in AI but working in that – suggestions for creating AI department in company? | Coursera Community
Coursera Header

New in AI but working in that – suggestions for creating AI department in company?

  • 11 October 2019
  • 2 replies

Badge +1
Hi, my name is Gabriel Santana, i am a math phd student from Venezuela. I've done many courses about data science and machine larning, recently i doing the Deep Learning course whit the excellent professor Andrew Ng and i was really fascinated and intrigated about the AI.

I'm working in a Company who oferte technologies services to airports here in Venezuela, the operative system that we use generate many operation data, and i uses that data for practice my skills in data science and i propoused a AI project.

With the Artifficial vision and the Natural Language Processing we'll create others dynamic system that show to ours clients all aspect with voice commands of their airports, from aeronautics operation and finantial data to security whit facial recognition and with that we want create the department of artificial inteligence in this Company.

Despite being new to this I understand the theory very well by my training. That is why I would like you to give me some suggestions for the success of this project.


2 replies

Userlevel 2

Hi Gabriel, 

Very interesting question actually. You mentioned that you did a deep learning course by Andrew Ng on coursera, I finished the deep learning specialization recently. He has a course titled “Structuring Machine Learning projects” if you have not gone through I highly recommend it. Professor Ng always stresses good practices when it comes to implementation and I believe that a correct methodology is very crucial to a good project. 

He and his team have also put together this fascinating study that starts with an overview of a project development lifecycle., I was not sure if I could send it directly to you but it is worth signing up for his website for it. 

With all that, you have asked how to set up a Department and I assume that a successful pilot project would help you get there, hence the above tips to get things going. Now, I don’t know how big or small your organization is, how much resistance to change (ie bureaucracy) you will encounter, how far along have you come in your project etc. - I really would need some more info from you on that, but from my experience, an AI initiative would not be any different from any other innovation.

I don’t want to go into details of setting up a proper project and organization as that will depend on the specifics of how things are done in your company but I can tell you about some hurdles that you need to keep in mind. For simplicity, I will go over troubles with data, technology and people.

  1. Problem definition is key - if you are not solving a business problem or need, then all the other steps won’t be necessary. Also, you need good criteria for what is success and ideally quantify it (90% accuracy, 80%?) If you are solid with this part, then remember the definition when facing hurdles, so as not to get sidetracked and lose sight of what is important.
  2. Usually, the data you are looking for will be disorganized, missing or some other form of “unsuitable” for AI. You will need to realistically look at options of overcoming this (how long it will take to process or generate the needed data)
  3. The second point leads to the question of tools - usually information is stored across several different systems, platforms etc. So ETL is required, it is up to your organization on how they want to do it (buy a vendor solution vs build one yourself), architecture, data pipelines etc. 
  4. People - think about how this department would look like, what kind of roles will you hire for and what kind of roles will be outsourced? Can some current people be trasferred over? What is your budget? What kind of projects will it undertake? Can you realistically get the right talent and give them interesting problems to solve for the foreseeable future?

I hope the above helps clear things up a bit, I wrote this post from my own experience being at both companies that implement AI as a customer and also as a solutions supplier, plus added my project management experience a bit. People (or you) might not agree with everything written above, but I hope at least we can start a productive discussion here. 

Also, the above is only a short list of things that need to be kept in mind and I do not delude myself that I have covered it all. 

Good luck!

Badge +1

Thank you for your suggestions.