Journey From Hindi Medium Schooling to Silicon Valley to AI Leadership
Hi All, I am Sandeep Singh, Head of Applied AI and Computer Vision at Beans.ai in Palo Alto, California, USA. I have over 19 years of experience in various software companies, including Misys, Oracle, Cisco Systems, Blackrock, and Beans.ai.
Background
I am from Prayagraj (formerly known as Allahabad), Uttar Pradesh, India. My father was an employee at NTPC, which allowed me to attend multiple schools across India. One thing that may not be obvious from my career profile is that my schooling until class 12 happened in a purely Hindi-medium school, and the only subject I studied in English was English itself. I had zero exposure to computers by then.
After completing school in 2000, I attended LNCT College of Engineering, a state engineering college in Bhopal, Madhya Pradesh, India. Following my graduation in 2004 from the computer science department, I secured a job.
The Journey from Bangalore to Silicon Valley
The college job offer was from Misys, a banking software company with locations in Gurgaon, Mumbai, and Bangalore. I preferred Bangalore because it was the IT hub.
After working for almost 2 years at Misys, I switched to Oracle and later to Cisco Systems. In 2010, while I was at Cisco, I received an opportunity to work in the United States, as they were inviting engineers from around the world. I was newly married at the time and had never visited abroad. There was some hesitation, but also an aspiration to see the world. It was a tough decision, but with the support of my family, we moved to the San Francisco Bay Area, popularly known as Silicon Valley.
With the goal of becoming wealthy and successful, I pursued a career in investment banking. To excel in this sector, I began taking the CFA (Chartered Financial Analyst) levels 1 and 2 courses. After 3 years of continuous studying and hard work, I received an offer from my dream company, BlackRock, for the role of an investment banking professional. BlackRock is a massive company that provides bonus amounts greater than the salaries offered by many other companies in the US.
First Exposure to AI
I live near Stanford University in the Bay Area of the USA. As you may know, the global excitement surrounding Deep Learning began with the design of the Alexnet architecture for image classification at Stanford University.
In 2012, even before the publication of the AlexNet paper, I could already sense the buzz and excitement permeating Silicon Valley about this breakthrough. People were constantly discussing it around me. This piqued my curiosity about Machine Learning and motivated me to complete the famous Andrew Ng course on Machine Learning. However, it wasn’t quite enough to inspire me to switch my career to the AI domain.
Motivation to Switch My Career in the AI Domain
In 2015, on a weekend, Andrew Ng and his team did a symposium, a conference including experts, at the Stanford Global Business School. It was all about AI, which was my turning point.
Only 200 people attended that, as the AI community was not so huge then, but that crowd included most of the AI rockstars of today's world. A few of them were Russ Salakhutdinov (Apple's AI head), Samy Bengio (Senior Director of AI and Machine Learning Research at Apple), Andrej Karpathy (Tesla AI Head), and Richard Socher (Stanford's AI Professor).
I was new to AI and didn't know how Deep Learning works. But after seeing all these people in that small room, I could sense the potential of this field. I am talking about that time frame when there was no Tensorflow or Pytorch, and people were using Matlab or Octave to build their ML solutions. Even then, people around me were keen to understand this domain's workings.
Money could keep you motivated, but not for so long. You need a much higher reason to be motivated every day. That weekend was transformative for me and inspired me enough to go ahead and resign from BlackRock. It was a crazy thing to do, but I still did that.
Master's Decision in Machine Learning, Deep Learning, and Artificial Intelligence
I left BlackRock and joined Oracle, near my place, to get more time to spend with my newborn. In parallel, I started reading books (although there were not many) and training myself in building ML models. I started with the book Artificial Intelligence: A Modern Approach by Peter Norvig (director of research and search quality at Google).
I also enrolled in an online part-time master's program at the Georgia Institute of Technology, US, for getting a specialization in Machine Learning, Deep Learning, and Artificial Intelligence. The intensity of coursework is amazing in US universities, and you need to work harder to pass these courses.
When I joined my master's, I already had a new baby, a very active social life, and a constant visit to my homeland India, but I was committed to my course. And the motivation I derived from the same two-day symposium. It was an online program but deadlier than offline courses as you must also take care of your other responsibilities. Somehow, I completed it successfully.
I met many different people across multiple industries silently inventing something that is not there in the world. I drew inspiration from them and grabbed the chance to know the wave of Deep learning. I was always keen to meet people doing the latest and greatest research.
In 2018, I got the opportunity to talk to Jeff Dean at Tensorflow.com, one of Google's most amazing programmers. While talking to him, I got to know that Google was already using more than 3000 AI models in productions. In 2017, Google launched the Google Translator, which was not so good during the launch, but within a week, it improved drastically. Jeff explained how they changed their traditional translation approach to NLP (Natural Language Processing) based translation. Even today, Google uses more than 10k AI models in production.
After Master's
I kept learning and doing AI after my master's, and slowly this journey led me to significant landscapes. I was always aware of what companies were building in the AI domain, and one of the primary reasons for that is I live near the area where everything starts (Silicon Valley). Intel, HP, Google, everything was founded in Silicon Valley. So I got to meet and learn from different people from the computer science industry.
I had an opportunity to meet and discuss with all the rock stars (mentioned above) of the AI domain, including Andrew Ng himself. I learned a lot of things from them and delivered solutions to my company which was impossible otherwise.
Joining Beans.ai as AI Head
Four and Half years ago, the founders of my current company, Beans.ai, Nitin Gupta and Aakash Agarwal, approached me with the idea of developing something big in mapping operations, and they wanted someone to fuse AI into it to do things in an automated way.
I joined them and started applying AI on satellite imagery from scratch. Since then, we built different models for Object Detection, Segmentation, and Classification tasks, along with many Machine Learning models for clustering, heuristic search, density estimation, and optimization problems. Recently, we built an automated way to check emails, read their content, extract the origin and destination locations, and automatically place orders.
Award Winner for Top5 GenAI Scientist
I got a chance to speak at many forums and conduct training for AI. Recently I visited Bangalore with a speaker and trainer role in the Data Hack summit organized by Analytics Vidhya. They felicitated me with an Award of Top 5 GenAI Scientist. I gave training on Stable Diffusion and text-to-image technologies to AI professionals in Bangalore.
Tips and Advice for Beginners in Deep Learning Domain
Just start your Journey and adopt the 'Learning By Doing' approach
The start of your journey does not relate to getting a job in AI. Even if you get a job in an AI company, your journey will start from that point. In the ML industry, new techniques are emerging daily, and you must keep yourself updated. Basics remain the same, but in production, one wants newer techniques that work the best, which are evolving very fast.
Learn Python
The concrete advice I want to give freshers is to learn Python programming. It has a similar magnitude of penetration in computer science to what Java had in 2004.
See AI as a technology, not an industry: The X + AI theory.
AI itself is nothing but a bunch of maths to perform backpropagation, detection, or classification. You need to consider AI a technology that must be fused with a domain. For example, Medicine + AI, Software + AI, Automotive + AI, Fluid Mechanics + AI, etc. You can be a doctor, lawyer, pharmacist, or be in any of these core domains and can still do AI to transform the way you do your core work. I believe "AI may or may not replace you, but a person using AI will replace you."
Adopt AI in the current company rather than thinking to switch into an AI Company.
Whatever domain you are in, try to adopt AI if you are interested rather than preparing to switch to companies working in the AI domain. If you are planning to be in this domain, then it's not always necessary to look for autonomous vehicle companies, Deepmind, or Tesla. Every aspect of your product or function in your current industry has the potential to use AI for simplification. Try to find that part and start applying AI there.
Keep yourself updated with the newer technologies in the AI domain
Keep yourself updated with the newer technologies that are coming into the market. Recent advancements in Generative AI in the text domain have opened the pandora-box, and a lot is doable with this trend, and many people have started doing that.
Go for a Post-Graduate Program in AI.
If your situation allows you to do a master's, go for it. In graduation, we only scratch the surface, but during master's, we actually see the depth of the domain.
Guidance for Interview Preparation in Computer Science and AI Domain
Practice DSA and System Design
No matter which branch you come from, what college you are from, or how long you have been doing software engineering, invest some time practicing Data Structures and Algorithms and understanding system engineering. By system engineering, I don't mean the hard-core implementation in C++ but developing the knowledge of Block diagrams of how components are placed inside the software, how to do load balancing, and all.
Practice questions from the 'already-asked' sections for the company you are targeting
If you are preparing for an interview for companies like Google or any other, go to the interview table and do all the problems previously asked in their interviews. Believe me that will be all to get you through.
When you start learning programming and doing Data structures and Algorithms, you are building the bottom layer of the pyramid. So you must be extremely honest with yourself. You shall ensure that you really get the concepts; otherwise, there is no point in showcasing the number of problems you have solved.
The same goes for Data Science or Deep Learning interviews, as it does not matter which college or branch you are from; you must know how to code; otherwise, you won't be able to experiment. It is the 0.0th part to crack any job in any fancy company working in the software industry.
Final Remarks
Where you come from has minimal relation to where you can go. If you know how a Markovian process in algorithms works: "What will be your next state depends upon your current state and the actions you are taking at your current state." It does not depend upon the history of your actions or the actions you have taken in your past. I believe life is similar to the Markovian process, and you can change your life anytime.
Enjoy learning, Enjoy algorithms!