Tushar Bhatnagar is an applied AI researcher and a serial entrepreneur. He has worked in a dozen small and medium scale businesses and startups and a diverse range of projects in healthcare, AR/VR, video content creation, autonomous systems, and deep-tech initiatives in the product management and data science space. In August 2021, Tushar joined Andrew Ng’s DeepLearning.AI as a mentor.
“Being a mentor is an amazing experience as long as you are devoted to your ideals. The whole experience can enrich your life on a personal and professional level and develop your ability to motivate and encourage others. This can further help you become a better manager, employee, and team member,” said Tushar Bhatnagar, co-founder and CTO at vidBoard.ai and co-founder and CEO of Alpha AI.
Sign up for your weekly dose of what's up in emerging technology.
AIM: Tell us a bit about yourself.
Tushar Bhatnagar: My goal since childhood was to create an impact. I like to call myself an innovator over a researcher. My vision is to take the research we do at Alpha AI to the masses.
To grow, one must try faster, fail harder, and create. I started my first venture as a sophomore. The notable ventures I have started include:
· INNVOS LABS
Later, I decided to get some experience in a corporate environment and took up a job as a Systems Engineer with Infosys. In a year, I found the tasks monotonous and not adding value to my career. Eventually, I decided to opt for a master’s in AI.
Then, I came back to India. While I was waiting to return to work with NatWest Entrepreneur Accelerator, Warwick Incubator and UCL’s CapitalOne, COVID-19 escalated in India, and I had to stay put. So I continued as a CTO in AIOTIZE, an AI-driven video analytics platform.
Later, I stepped down as CTO and decided to start Alpha AI. Besides Alpha AI, I am also the co-founder and CTO of another deep tech startup called vidBoard.ai.
We look forward to revolutionising the way we see our world by building safe AI systems to solve modern problems. All our research projects are simple yet unique with patent-pending technologies.
AIM: How did your fascination with AI begin?
Tushar Bhatnagar: In 2015, I was trying to design UAVs. While trying to develop “out of the box” solutions for the problem, I came across artificial neural networks. And I was hooked. I read a lot of research papers to understand things better and know more. One of the papers–Deep Learning by Yann LeCun, Yoshua Bengio and Geoffrey Hinton–helped kick off my AI journey.
I started wondering, “what on earth is AI?” and why is it hard to make things work using ANNs. As time passed, the whole AI space pivoted to using Python-based frameworks for designing and implementing AI algorithms, and I was able to design systems leveraging the frameworks from Google and Facebook i.e., TensorFlow and PyTorch.
AIM: What are the most interesting projects you’ve worked on as an AI researcher?
· Book Study Room using Amazon LEX: Designed a chatbot that uses NLU and NLP to help students book study rooms. The bot looks for available study rooms and matches them with the student’s request on a priority basis.
· QuickKart: Automated booking, route planning and multi-modal journey planner for the travel industry.
PayGo: We wanted to design a system similar to Amazon Go but hardware agnostic. The algorithm helps manage inventory and automate the payment process– all through computer vision.
· vidBoard.ai: It started as a solution to generate synthetic talking avatars for educational content. Later, we converted it into a full-fledged tool to generate video presentations at scale with marginal costs.
· LawDiktat: I always wanted to develop a product that could reduce the turnaround time of paralegals, advocates and even judges when it came to documenting understanding, due diligence, case prediction etc. The project has now been scaled into a commercial product.
AIM: Tell us about your AI startups.
Tushar Bhatnagar: When was the last time any traditional or small business benefitted from AI research and technologies? Alpha AI is trying to bridge that gap. We are a research-driven organisation aiming to be a global catalyst for promoting scientific research and building products for public benefit.
We aim to address three critical problems:
· The lack of democratisation of AI and its slower adoption rate across public domains and SMBs.
· The gap between traditional businesses, scientific or industrial research and academics.
· We want to help students and entrepreneurs design and deploy AI systems that promote scientific discoveries and solve real-life problems.
With vidBoard.ai, we want to generate short movies within minutes. Our tool uses GAN-based networks and speech synthesis models to create talking avatars. The pipeline will be available to the public in Q1 of 2023.
AIM: How did you become a mentor at DeepLearning.Ai?
Tushar Bhatnagar: From 2016 to 2021, I have been actively participating in a lot of Q&A sessions and pursuing the courses offered by DeepLearning.AI.
Simultaneously, I had been doing certain courses by Udacity and Nvidia DLI to strengthen my core knowledge of AI. Being a mentor was not an easy journey. I usually spent 6-8 hours a day going through various courses and 4-5 hours a day writing code and doing assignments.
The DeepLearning.AI team reached out to everyone who completed all the specialisations with an opportunity to be a mentor. I applied, and after an initial round of screening, I was made a part of the mentorship team for AI in Medicine.
As a mentor, I must address all the queries from the students who have taken the course assigned to us.
AIM: What are the must-have skills to build a promising career in AI?
Writing software: Data scientists oversee designing, developing and enhancing software-based systems and developing automation tools to help them run more efficiently. Practical knowledge of Python, R, Java, and SQL are essential skills. In addition, they should have strong software architecture and development skills.
Playing with data: Data scientists must know how to extract meaningful information from data and work with enormous datasets. They must know how to convert raw input data into clean and accessible datasets to train machine learning models. They must also be familiar with the databases used to store and deliver data to the machine learning system. The ability to manage SQL and non-SQL databases and employing big data tools like Hadoop are highly desirable
Data scientists should know the fundamentals of algorithm design, statistics, etc. Those who work in machine learning must be familiar with the mathematical and statistical ideas that underpin AI and ML. They must be conversant with AI and machine learning algorithms and architectures, particularly deep learning.
Soft skills: You need good communication skills to translate business requirements to machine learning terms. You should know how to organise and manage projects for system implementation and possess strategic vision and team-building skills to collaborate with coworkers and customers effectively.