An interview with Alessio, Lead Data Scientist at FloydHub
Who’s Alessio?: Our interviewee today is a developer, computer scientist, and entrepreneur. He’s also FloydHub’s Lead Data Scientist. As with every startup, he wears a few other hats, too, and you can also find him on Customer Support, Product Development, & Content Marketing. You can learn more about Alessio here.
I’ve worked with Alessio on a number of articles I wrote for FloydHub, and he’s always been helpful and motivating. I was so excited about the chance to interview him and I hope that our conversation is as inspiring for you as it was for me. In it, we talk advice for the job market, what drives us to work in data science, and what it’s really like working at a startup that’s trying to change the way the world builds and uses AI. I hope you enjoy it!
An interview with Alessio, Lead Data Scientist at FloydHub
Sayak: Hi Alessio! Thank you for doing this interview. It’s a pleasure to have you here today.
Alessio: Hi Sayak, and hi to all of you that are reading :) Thanks for having me.
Sayak: Maybe you could start by introducing yourself — what is your current job and what are your responsibilities over there?
Alessio: I’m currently working as a data scientist at FloydHub, and on any given day, I’ll be:
- leading data science, which means: set experiments in order to validate different hypotheses we have, derive useful insights from data we collect, test new methods/techniques to help our customers, and so on. Let me try to reframe this with a seafaring analogy: I’m like the officer navigator who helps the captain (CEO) to shape the course in the sea of uncertainty (market) that our ship (startup/company) is sailing.
- helping with product development and applying our data-driven framework, the Product Improvement Engine
- helping our clients, who are a mix of startups, researchers, & established companies with technical customer support
- managing our content strategy and our blog, whose goal is to distill AI knowledge and contribute to a collective understanding
For those of you familiar with startups, you know that’s a pretty typical grab-bag of responsibilities! I knew what I’d be getting into joining an early-stage startup (thanks to Paul Graham’s essays and YC), and FloydHub has never failed to provide me with amazing challenges and opportunities for impact.
Sayak: That is wonderful. It must get sometimes very challenging. But I hope it’s worthwhile in the end. How did you become interested in data science and machine learning?
Alessio: I can tell you the exact moment, actually. I was in an SWE seminar during my last year of studying CS at Bologna University.
The researcher had introduced an interesting problem: finding a way to automatically classify the emails of the Linux Kernel Mailing List. They walked us through a machine-learning solution, using Unsupervised Learning, since it used K-Means clustering, to complete the task, and I fell in love. The concept that a machine was able to automatically solve a task without being explicitly programmed to do so…well, it struck more than a chord in me.
Even today, years later, whenever I think about Differential Programming (or Software 2.0 or whatever naming convention you want to use), I get super excited about the potential that this technology has and the impact it will make in the very near future.
That’s really why I decided to join FloyHub — their CEO’s mission resonated with me. “Helping people to build impactful AI applications.” That’s exactly what I want to do with my career, and so far, it’s been an amazing, rewarding challenge.
Sayak: So wonderful! And it really pleases when I hear that perspective from Sai (Sai is the CEO of FloydHub). When you were starting what kind of challenges did you face? How did you overcome them?
Alessio: I’ll answer your question in the two ways I think will be most helpful.
First, I’ll cover the challenges I faced in looking for a job. I knew I needed to show something concrete to my future employers, so I enrolled in courses (Caltech, Andrew Ng, etc…) while I worked on my own startup. For many reasons, mostly due to a lack of management skills, I gave it up after about two years and started looking for a job.
After a couple of applications, I realized that my CV wasn’t good enough to capture the attention of any of the big tech companies (FAANG). So I turned to Plan B: joining an early-stage startup and helping them to succeed.
It’s far easier to get an interview with startups, and the process tends to be more efficient than at big companies, less broken. If you’re motivated and a good interviewer, you should have no problem getting job offers. If you’re finding yourself in a similar situation, I strongly encourage you to apply to startups. I’m planning to write a blog post about this, but until it’s up, I’ll leave you with this great article: Why You Should (And Shouldn’t) Join a Startup.
Second, I’ll talk about being a data scientist in a company.
Data science is an emerging profession. The boundaries of the job are not clearly defined by the industry yet; each company is trying to figure out how to really use AI. There are tons of challenges in terms of knowledge and training gaps at the executive/product manager level as well as no consensus on processes, tools, or best practices across companies.
It’s both interesting and hard to stay on top of a still-evolving field. There are only two solutions I’ve discovered: experience and asking for help. Experience helps you develop your intuition, and the process of gaining it can be made easier by leaning on other people who’ve done this before. At FloydHub, oftentimes we work with companies who are just starting down the AI path. Our goal is always to be true partners and to help those companies figure out how to transform their business with AI; we sit down with executives and data scientists and show them how to think about their AI strategy and the low-level implementation details. Within time, they’re finding success and we can transition to less of an active role.
So to sum up: try to figure out as much as you can and seek out help when you get stuck. There’s plenty out there, and our blog is a good place to start.
Sayak: That was so nice of you sharing the views in such a detailed way, Alessio. You put out two ways which I think are really interesting. That particular part where you mentioned about the less broken nature in startups. I strongly agree with that. What were some of the capstone projects you did during your formative years?
Alessio: Even if my startup idea never saw the light, it was such a great learning experience; managing an ML project from scratch gave me a lot of self-confidence in how to approach that type of task. Every time a FloydHub client needs support or one of our internal projects needs management, I apply what I learned back then. Contributing to Open Source ML/DL project is definitively second on my list of “very much worth it” projects to experiment with. It’ something that I strongly encourage you to do. Don’t fall into “I don’t have a PhD in ML, I can’t do this!” imposter syndrome trap. Of course you can!
Sayak: Thank you for sharing that, Alessio. These fields of data science and machine learning are rapidly evolving. How do you manage to keep track of the latest relevant happenings?
Alessio: This is a great question! I’m still trying to figure it out the right approach, but this advice, which comes from HackerNews user mlthoughts2018, is some of the best I’ve come across:
1. Read software libraries first. If new techniques make it into standard implementations of major libraries, it’s a solid signal the new technique is actually worth your time.
2. Don’t bother with Arxiv / conferences / etc., until a paper is at least a year old and remains highly cited & investigated. In DL especially, papers are often slapdash permutations of existing ideas applied with one weird extra trick that ekes out “SOTA” results in one benchmark. Architectural choices are often not motivated by any evidence, and whatever scramble of grad student descent was needed to arrive at the final architecture choice is conveniently left out of the paper (and often would call the result into question from a multiplicity of testing POV).
3. Read what you enjoy. It shouldn’t feel hard to stay current in your main area. If it is feeling hard, you’re trying to force it. Maybe pick a subspecialty you enjoy and only focus on that for a while.
4. Forgive yourself for not staying current. You are busy, reading research papers takes time & is not often the best use of time compared with competing interests. This is normal. If some situation/job interview/academic program is making you feel otherwise, it’s probably a red flag about that thing.
Sayak: I will be forever grateful for this gold, Alessio. I am sure the community is really going to benefit from this a lot. Being a practitioner, one thing that I often find myself struggling with is learning a new concept. Would you like to share how do you approach that process?
Alessio: I totally agree with Dr. Rachel Thomas on this: you (yes, you) should blog! Learn by doing (in this case, writing and coding) is the best way to fix new knowledge in your mind. And it has other great benefits, like giving you a way to show your value. We’ve started a program at FloydHub called the FloydHub AI Writer program, whose purpose is to help develop the ML/DL community by giving them a platform to share their knowledge and in turn to give back to that community by providing top-quality, free tutorials and guides to improve or start their data scientist careers. I’m super satisfied with how it’s going so far, as are our repeat writers. Emil was recently featured on Google for this article, you, Sayak, are now doing events around the content you’ve created for the blog, and Cathal transitioned from SWE to MLE at Intercom. There are many more success stories and we’ll soon be sharing them on a dedicated page on FloydHub Blog.
Sayak: That’s pretty exciting, Alessio. For me, joining the FloydHub AI Writer program was a huge turning point in life. I get to learn and brainstorm so much that I have gradually become addicted to it. Any advice for beginners?
Alessio: Here’s my advice:
- Get out of your comfort zone, but not too much. Learning from MOOCs and getting certificates is good, but it’s really unlikely that that experience alone will help you get an ML/DS job. Work hard in a field where you can move around and you are sincerely passionate.
- Study the basics of Computer Science (CS)! Seriously, this is extremely important and too often overlooked. DS is built on top of CS; if you miss this piece, it’s like building a house without a foundation.
- Your ego is the enemy. The sooner you will realize that, the better it will be for you. There is a terrific article on getting over the fear of failure and the embarrassment of being a beginner by Greg Brockman that I highly recommend you read
- Be a lifelong learner. This skill will help you survive in the job market as well as the real world and will make you a more interesting person along the way.
- Set a goal & enjoy the journey! I think that Monster University ending is the perfect description of this: https://youtu.be/S3zW6mHlues
Sayak: I am sure I am taking all of the advice because I think it’s all simply gold. Thank you so much, Alessio, for doing this interview and for sharing your valuable insights. I hope they will be immensely helpful for the community. How do people get in touch with you for advice?
Alessio: Thanks, Sayak, for this opportunity! I really hope that my experience will help other aspiring data scientists find their path. At Floydhub, we work with companies and the community through our office hours program to share best practices. Email us questions, ideas, or things we should read: firstname.lastname@example.org.
I think it is very important to see through one major point in this interview — how Alessio’s dauntless hard work allowed him to shine. How Alessio focused on developing his craft tirelessly is something really inspiring. Alessio shared a number of extremely important tips throughout this interview which I am sure will be tremendously beneficial for the community. Being a life-long learner is what keeps going in a better and makes us survivable in this competitive market. I think this is equally applicable to beginners as well the experienced professionals in any field.
I hope you enjoyed reading this interview. Watch out this space for the next one and I hope to see you soon.
If you want to know more about me, check out my website.