We recently connected with Vaisakhi Mishra and have shared our conversation below.

Vaisakhi, thanks for joining us, excited to have you contributing your stories and insights. So, let’s start with trends – what are some of the largest or more impactful trends you are seeing in the industry?

Oh, what an exciting, albeit a little scary and uncertain time to be working in Data Science and Artificial intelligence. In 2023 my husband and I (him being a crazy gamer and me catching on to that craze because of COVID) were building a gaming rig, and at some point, we wanted a specific graphic card. Chat GPT had released a few months back and we initially thought the hype around ChatGPT will die down just like it had for VR and AR some years back, but when we were searching for that graphic card, it was out of stock at every electronics store around us and I had a feeling this time this technology hype is perhaps here to stay. Fast forward to today, AI has become the term of the year, with every other company talking about it, figuring out where it fits their needs, trying to make AI a part of their day to day somehow so that they don’t miss the train, getting excited and overwhelmed with the pace at which the technology is evolving and flooding the market – and as an AI Engineer, I am at the center of this storm right now.
So, if I had to talk about the trends, there are just way too many to talk about in so many different branches and sub-techs that makes up the current AI world. Just a few years back industries hesitated to adopt AI, because it felt complex and out of reach, but now everyone wants a piece of the pie, not because they complex part has changed, but it seems way more accessible and is evolving so fast that it has taken over everyone’s thoughts. Last year LLMs aka large language models were all the rage, but now even those models are evolving. AI systems are increasingly integrating diverse data modalities (e.g., text, images, audio, video) to enhance their contextual understanding and application, which is exciting to every kind of business. Models like GPT-4V(ision), Gemini Ultra, Llama Vision and CogVLM exemplify this trend by seamlessly combining modalities for tasks like visual question answering and real-time processing. Plus, we are seeing an advent of a sea of more specialized models – finance, law, health, code, so many. Experts in the field also are turning their focus towards building specialized language models and multi agent systems to have meaningful assimilation of AI technology in their businesses. The potential and reach of AI now is just massive.
And while the opportunities are aplenty, the ethical concerns of what the models are trained on, the copyright infringements, the lack of visibility and explainability of models, the legitimate concern of lack of hardware accessibility and high resource usage and carbon emissions, the fear of loss of jobs because of AI automating certain parts of a job, are all there and the fear of AI around these topics is justified.
From a personal experience, I remember building an elaborate assistant for a client in 2023 using a 70 billion parameter LLM, but when they decided to deploy it for their clientele, the resource consumption and cost associated with it was something they just could not justify. And when one of the executives asked me about biases within the model, I had no answer, since not everything in the model was transparent or explainable. We did a second iteration where tasks were divided with smaller specialized and fine-tuned models with some validation testing to understand training data biases, which made the assistant a little transparent and significantly reduced the initial cost but made the assistant a little bulkier to maintain. After a lot of back and forth and convincing, the client finally rolled out this new assistant but now has come back to us with a request to enhance the assistant again with multimodal models and reduce the assistant complexity, not caring anymore about the training data – just because it is the new thing in the space. And frankly the complexity can be reduced now with so many agent frameworks available, so their ask is not completely unfounded – but they just deployed the assistant a few months back. Basically, it is all moving too fast and no one wants to be left behind.

Awesome – so before we get into the rest of our questions, can you briefly introduce yourself to our readers.

My name is Vaisakhi Mishra, and I am a Data Scientist and an AI Engineer, currently living in New York. But I come from a city called Cuttack in Odisha, India, lived a nomad life in India before moving to Seattle USA for my Master’s at University of Washington and then NY for my work at IBM. Since childhood, I have always loved finding patterns in information, curating things the best I can, associating one thing with another, until the “Eureka!” moment hits in my brain lighting up a bulb. I always used to be good at quizzes in school, and at debates because of this skill. In general, I am very curious about our world –systems, history, nature, data.
Growing up I had two dreams, working on something (a job) that would keep me excited and traveling (exploring the world). Exploring the world was something I was able to do as soon as I graduated high school and sort of did in part even before that because we moved around a lot within India. And since I loved journaling my travels, I got into travel blogging, which I have no plans of stopping. But the job part was a totally different story. I always saw my parents talk about their jobs as a necessary evil. They worked full-time jobs but never seemed to enjoy their work. I had seen them struggle and be unhappy and downcast because of their jobs as a kid. I didn’t know what I wanted to be until I decided to leave a stable job in India and started my Master’s in Information Management the US. Data science tickled my OCD for organization and discovering stories and patterns in data and I decided to become a Data Scientist.
Eventually the intersection of my academically acquired skills with deep learning and AI engineering, and the application of AI in solving real-world problems grew, I realized I was already solving such problems in my current day to day life, and AI Engineer was just a new title with the same exciting responsibilities. So here I am now, an AI engineer navigating the ever-changing landscape of AI technology and helping others do the same. As part of my job, I get to have real social impact on day-to-day lives of various strata of people, including the underprivileged. In one such project, I was able to use my AI skillset to mitigate adverse changes in neighborhoods due to gentrifying and other economically declining factors. I was extremely proud to be a part of the team who worked on this. Another project that turned out to be extremely impactful was when I was working on finding new and innovative way to detect and mitigate cyber threats in real time. I also mentor aspiring data enthusiasts of my Alma Mater and collaborate on research projects with open scientific collaboration communities. As you can guess now, my work is never restricted to one kind of problem or industry and this keeps me excited – aka the kind of job I always wanted growing up.

What’s a lesson you had to unlearn and what’s the backstory?

As a fresh graduate, I was always pursuing perfection in the software code I wrote, and this was a hindrance to progress I could have made on a day-to-day basis and other learnings I could have had chasing progress instead. I struggled quite a bit because of this during my first job. I worked on some 10-15 year old legacy code and the struggle to maintain and redo it always drove me crazy, so perfection became my objective. But over the years, in every project I was on with any client, I always found out that time was a stronger driver and constraint than perfection. It was very rare for perfect code to be a requirement. I also realized that perfect code doesn’t exist, especially not on the first try. This was, in general, something I took back as a life lesson and making progress on everything I was trying to do, every day, was better than chasing perfection in that thing. You don’t get from 0 to 100 directly, and it is ok, maybe it isn’t meant to be that way. Work out all the numbers in between and maybe something works. This also made me realize, never be afraid to take on a new challenge because of the fear of failing, it is ok to fail sometimes because might just accelerate your path to success the next time.

Are there any books, videos, essays or other resources that have significantly impacted your management and entrepreneurial thinking and philosophy?

The books that have significantly impacted my management and life philosophy are the Art of War by Sun Tzu, From 0 to 1 by Peter Thiel and a few biographies of notable founders such as Jobs by Walter Isaacson and Shoe Dog by Phil Knight.
There are several life lessons in management, creativity and innovation that I take back from Steve Jobs’ biography. He always thought people have ideas but a little creativity and a little innovation with the right direction is what makes them see the right idea, or the thing they didn’t know they wanted, because they had never thought of it. Even in From 0 to 1, Thiel emphasizes on creating something truly unique and valuable and not just reinventing the wheel or following the crowd, because most people will do that if everyone around them is doing it. This I think has always been a huge driver in how I approached new ventures or research, and it is interesting how this is so relevant now in my day-to-day work. People think of AI day in and day out now, try out so many interesting or even funky things, all powered by AI, that frankly they had not thought of earlier, or thought possible before now, they want to harness the power of the technology – but struggle with the ‘why’ and ‘how’. I am glad to be working in the field with expertise to help people realize the ‘why’ that they might have not even thought about and nudge them to use the newest and right technology for their needs as the space evolves at lightning speed, I actually get to turn the hype into a concrete and unique need of the hour solution.

Contact Info:

• Website: https://thewordpool.blogspot.com/
• Instagram: https://www.instagram.com/vaisakhi/
• Linkedin: https://www.linkedin.com/in/vaisakhi-mishra/
• Other: Google Scholar – https://scholar.google.com/citations?user=9XuWbQ4AAAAJ&hl=en