Nowadays, our homes are full of various smart devices, from a thermostat to phone to television.

It is not a surprise that most of the applications running on our devices require some sort of machine learning models to provide more comfortable user experience. Moreover, there is a growing need for a single application to run multiple models simultaneously.

For example, a safe navigation application, which based on the order given by a user will tell the direction using a camera, may need two models: YOLO (for objection detection) and any natural language model to understand the user's text/voice.

Now, in a scenario of a smart home, when multiple devices are running a variety of models, it is certainly the case that there is an overlap among all AI services. For example, both TV and a phone will run a natural language processing model to allow a user to voice their request instead of typing. These two different devices run the same model, which results in a redundant computation and a waste of energy.

What if it was possible to develop a system for closed environments such as a home or a university campus, which will provide AI as a service. In other words, what if it was possible to decouple the machine learning models from applications and applications from the devices? In this case, the applications can provide the same AI features while avoiding the need to perform heavy model inference on multiple devices.

This work touches mainly on the fields of Federated Learning and Edge Computing. This idea is quite feasible to achieve. I'd love to know your take on it.

Open Data Challenge will take place in New Uzbekistan University in January. I got the second place in 2020, four years ago. It’s a great opportunity to learn and create something meaningful. I highly recommend you to join if you have a chance.

Let’s return the Witcher reaction back. https://t.me/boost/tenscores

To uphold our traditions in the upcoming winter break, George and I are planning to do the following challenges:
1. Solving 50 Leetcode Medium problems in a day.
2. Create a music album in a day.

P.S. Some of you might know that I am a big fan of rap. Only a few know, though, that George and I actually created a couple of songs back in our freshman year. This might sound as weird and unexpected from me, but I want to try something else than coding. Everything as usual will be streamed on my YT channel.

When out of nowhere you get two days of break

One advice that I keep hearing from Computer Science seniors and graduates is that school is important, yes, but we need to be smart about our career and future plans.

Since the start of this semester, I have been thinking of what I want to pursue in the future and whether I need to go to the industry in the first place or continue with academia since it feels safer.

For long, I have been prioritizing my studies over any other thing in life. It worked well, but I have come to a realization that this is a losing strategy.

For long, I have been making excuses not to make time for other incredible opportunities and projects that I am able to do if only I have some free time during the semester. I also realized that this is never the case. I would never be actually free if I keep this mindset.

It’s more about setting the right priorities and making the time to focus on them. It’s about removing distractions and focus on what’s actually important.

As of now, I still find myself spending all of my time for courses. This is good in terms of academics, but not smart in the long-term perspective. I have only 5 semesters left till I graduate, and I need to use them more wisely.

Is there anyone planning to do PhD in Computer Science? If yes, what area do you want to pursue?