Invaluable tips from a Senior Data Scientist - DSBoost #1

Welcome to the first issue of DSBoost, the weekly newsletter where you can discover interesting people in the ML/AI world, get the main takeaways of a relevant podcast and stay up to date with the latest news in the field!

💬 Interview of the week

This week we interviewed Akshay, who is a senior Data Scientist at TomTom. Enjoy:

• What did you study?

  I have done a Bachelor of engineering in Electrical & Electronics engineering & a Masters in Mathematics. I have a background in programming & mathematics.

• What are your favourite resources?

  • I like Coursera a lot.

  • Made with ML for MLOps & Applied ML stuff.

  • ISL is one of my favourite books.

  • I also highly recommend following Andrej Karpathy’s work for education freely available on YouTube!

• What got you into your current role (portfolio, certification, etc.)?

  • I was fortunate to get a great mentor when I started.

  • I started as a computer vision intern & worked on a product for BMW.

• What do you enjoy the most in your work?

  Finding simple & effective solutions for hard problems!

• What tools do you use the most / what are your favourite tools?

  • TensorFlow, PyTorch, Pandas, NumPy are some of the libraries I use almost everyday at work.

  • Lately I have been using Microsoft’s Azure (Azure ML workspace in particular) a lot.

  • I’m a longtime PyCharm user (hoping to make a switch to VsCode soon :) )

• Do you use ChatGPT or other Al tools during your work?

  • I use ChatGPT whenever I can for generating code, solving errors, creating docstring & even for writing unit tests.

  • It certainly has a great impact on productivity if used properly.

• What is your favourite topic within the field?

  • I work for a company that builds maps & works on location intelligence. Everyday we deal with digital road networks & its associated data such as speed limits, turn restrictions & places near the roads. This data is inherently a Graph. I have used graph neural networks to solve certain problems where traditional methods failed. This is the reason I like GNNs a lot.

  • If you are looking to get started check StellarGraph python library & Petar Veličković’s youtube lectures.

  • Apart from that I like Transformers a lot due to the simplicity of their design & effectiveness of the attention mechanism. 

• Which one of the recent AI/ML models will have the most significant impact on the industry in your opinion?

  The answer is simple LLMs.

• What are you currently learning or improving (topics you are interested in nowadays)?

  Currently I’m trying to learn and adopt best MLOps practices.

• What is the biggest mistake you've made?

  • We should do one thing at a time, at times while trying to ride two horses we achieve nothing.

  • Starting is the hardest thing. Many times I have failed to work on ideas that I have in mind, partly because the tech world moves swiftly & it becomes hard to prioritise & focus.

• What is your biggest achievement?

  I have three patents in the field of Machine Learning.

• Can you share a fun fact about yourself?

  • Usually coders like working late hours, I’m a morning person.

  • Apart from work & coding, I’m an avid sports fan. I watch & play tennis and football on a regular basis.

🎙️ Podcast of the week

Ken's Nearest Neighbors - The TRUTH About Landing Your First Data Job (Avery Smith) - KNN Ep. 132

Key takeaways:

1. Formal education is not always necessary in data science, and online learning platforms make it easier to gain the skills needed without significant costs.

2. When hiring for entry-level data roles, it's more important to look for qualities such as willingness to learn, problem-solving skills, and real-world experience, rather than just a degree or specific technical skills. Training and professional development opportunities are also crucial for entry-level employees to grow and develop.

3. Self-motivation and discipline are important for staying accountable to oneself and others when pursuing a new skill or career path. Setting clear goals and processes is essential to stay on track.

4. Experimentation with different projects and tools is important for gaining a deeper understanding of the field and finding a unique path. It's important to take inspiration and ideas from others' work but put your own twist on it to create something new and unique.

🧵 Featured threads

Tivadar Danka @TivadarDanka

The expected value is one of the most important concepts in probability and statistics. For instance, all the popular loss functions in machine learning, like cross-entropy, are expected values. However, its definition is far from intuitive. Here is what's behind the scenes.

Kareem Carr | Data Scientist @kareem_carr

DATA SCIENCE CAREER ADVICE: Read this if you're interested in a data science career but feel like you might not belong.

🤖 What happened this week?

1. OpenAI’s CEO, Sam Altman, in his latest interview said that the release of further GPT versions will be slower than people expect. This will disappoint a lot of people, however, he thinks that they will appreciate it in the end, as OpenAI needs to make sure that this is done safely and responsibly. He didn’t hint whether GPT-4 will be multimodal or not, however, he suggested that it is to expect in the near future.

2. deeplearning.ai released a new course oriented to those getting started in the field. Its name is “Mathematics for Machine Learning and Data Science Specialization“ and it covers the fundamental mathematics toolkit of machine learning: calculus, linear algebra, statistics, and probability.

👥 Under the radar

Here are a few words from Sid personally:

I am Siddhesh Bangar, currently pursuing my undergrad. I Also work as an AI Researcher. Being a Data Science practitioner I love sharing my knowledge as a content creator in the form of Twitter threads and blog articles.