Computational Neuroscience, in many ways, will help in advancement of AI based technologies by enhancing our understanding of deep neural networks in extracting meaningful patterns from data.

Saurabh Khemka, our next pathbreaker, is Head of AI at Parspec, a startup focused on applying Machine Learning models and Generative AI technologies to Construction Tech.

Saurabh talks to Shyam Krishnamurthy from The Interview Portal about his fascinating work on brain-computer interfaces (where he applied machine learning to brain signals ), and how computational techniques can be used to understand the brain and human behaviour.

For students, the only way to learn is to work on real problems—freelancing, open-source, hackathons—anything that pushes you out of theory and into messy, practical problem-solving !

Saurabh, can you share your background with our young readers?

My name is Saurabh Khemka, and I come from a small village in Bihar, India, with a population of fewer than 5,000 people. My childhood was simple, surrounded by rural landscapes and limited resources. I studied in a government school within the village. I still remember the countless nights spent studying under a lantern, with a dream in my heart to learn maths and physics—subjects that fascinated me even when I had little exposure to the outside world.

My father is a small business owner, running a clothing store in our village. My family’s values around resilience and hard work had a profound impact on me. Until class 10th, I had never stepped outside my village, but I knew that to pursue science, I had to go beyond my comfort zone. After 10th, I moved to Bokaro for my higher secondary studies.

The transition from the Bihar Board to the CBSE curriculum was challenging. I failed twice in midterms and experienced academic failure for the first time. But that period of struggle shaped me. I worked hard and ended up as one of the top scorers in my school. That’s when I first heard about the IIT JEE—and realized how it could change my life. Knowing Bokaro wouldn’t provide the right preparation, I took a leap of faith and moved to Kota for coaching. That decision changed everything. I eventually cracked IIT.

What did you do for graduation / post graduation?

I pursued my BTech in Biotechnology from IIT Roorkee. This was another major turning point in my journey. Despite coming from a modest academic background, I consistently topped my department, thanks to the supportive professors and academic culture. Outside the classroom, I found joy in playing snooker, which helped me stay grounded and balanced.

My academic success at IIT gave me the confidence to apply abroad, and I was fortunate to receive a full scholarship from the Swiss government (granted to only a select few in India) to pursue my Master’s at EPFL, Switzerland. It was at EPFL that I was first introduced to machine learning and robotics, which deeply fascinated me.

This interest led me to pursue a PhD in Neuroscience at the University of Zurich, where I focused on computational modeling and continued working with deep learning and ML frameworks.

What were some of the key influences that led you to such a unique career in AI & Machine Learning?

Several influencers and turning points shaped my decision to pursue this field:

Key influencers: My own curiosity about how the brain works and how machines can mimic intelligence. The scientific rigor I encountered during my master’s studies further deepened this interest.

People/Mentors: My professors at IIT Roorkee and EPFL played a critical role in encouraging me to explore research, always pushing me to ask deeper questions and think independently. Later, during my transition to industry, my manager at Ittiam Systems became a key mentor—helping me understand how to apply research in a real-world, product-driven setting and guiding me in navigating the initial phases of my industry career.

It goes without saying that family support (parents and siblings) has been paramount in whatever I’ve achieved so far. My wife Puja, whom I met during my time at IIT Roorkee, has been a consistent source of strength and encouragement—especially during difficult transitions and moments of doubt. Her belief in me and unwavering support gave me the clarity and energy I needed to keep moving forward.

Events: Exposure to real-world applications of ML during my research years helped me realize I wanted to build solutions rather than just write papers.

Turning points: The decision to pursue a PhD, and later the realization that I wanted to transition to industry, were both pivotal in shaping my current path.

How did you plan the steps to get into the career you wanted? Or how did you make a transition to a new career? Tell us about your career path

My academic journey has definitely been non-linear — moving from Biotechnology to Bioengineering and eventually to Computational Neuroscience. But each transition was a deliberate choice, aligned with my evolving interests in how data and technology intersect with biology.

When I was considering my options for a Master’s, I had the opportunity to choose from several reputed universities — including some strong programs in the US, the University of Toronto, and a few across Europe. I ultimately chose EPFL in Switzerland for a very specific reason. Although my major was in Bioengineering, EPFL offered incredible flexibility: I only needed to take a few core department credits, and the rest could be selected from across other disciplines. This allowed me to shape a highly interdisciplinary program tailored to my interests.

I took courses in data analytics, brain-computer interfaces (where I applied machine learning to brain signals to control devices), robotics, and nanotechnology. This exposure to both applied machine learning and emerging tech sparked a deep interest in research — particularly in how computational techniques can be used to understand the brain and human behavior. That played a significant role in shaping my decision to pursue a PhD.

Although I often list it simply as “Neuroscience,” my PhD was specifically in Computational Neuroscience, which is quite different from the more traditional, medically-focused neuroscience paths. My work wasn’t clinical; instead, I was working with brain signal data — primarily from MRI and MEG — and applying statistical and machine learning models to extract meaningful patterns from that data. The goal was to understand how the brain encodes and represents cognitive processes under various conditions.

In fact, I was a certified MRI operator as well — a fun but important part of the work — since I was actively involved in the collection of neuroimaging data from human participants. The hands-on nature of both data collection and modeling made the experience incredibly valuable.

So while my background may look non-linear on the surface, there’s a consistent thread: using quantitative and computational tools to make sense of complex biological systems. That has continued to be a central theme in my career ever since.

My career has been a journey of continuously adapting and learning: I’ve always believed in long-term goals with short-term adaptations. When I decided to move to industry from academia, I knew I had to upskill.

Internships/Jobs:

  • During my PhD, I freelanced on ML and statistics projects to gain real-world experience.
  • My first full-time role was at Ittiam Systems Pvt Ltd, an Indian company where I worked as a Lead Engineer on research-based computer vision applications for edge devices.

Professional networks, mentors from academia, and peers in industry helped me transition smoothly. LinkedIn and freelancing platforms also helped me stay engaged with real-world problems.

After my time at Ittiam, I joined Philips because I was keen on applying machine learning in the healthcare space. It was an exciting opportunity to contribute to a meaningful domain. However, I quickly realized that healthcare is often a slower-moving industry, especially in terms of innovation cycles and deployment. I was looking for faster growth and more dynamic challenges, which led me to transition to Walmart.

At Walmart, I found what I was looking for — large-scale data problems, rapid experimentation, and the chance to work alongside some of the brightest minds in the industry. I spent over three and a half years there, working on everything from customer behaviour modeling to personalization and fraud detection. It was a rich learning ground in terms of both technical depth and business impact.

After that, I moved to MFT Energy, a trading firm. This shift might seem unexpected, but it was driven by a personal connection — one of my clients from my earlier freelancing days offered me a great opportunity. I was also genuinely curious about trading and wanted to explore how data and machine learning could transform decision-making in that space.

Throughout my career, I’ve intentionally chosen to explore different domains. My focus has always been on solving challenging problems using machine learning, rather than confining myself to a specific industry. I’m an explorer by nature — unafraid to take on new challenges and venture into uncharted territories — and I believe that mindset is reflected across my journey. While not every role was directly tied to my PhD, the analytical thinking and modeling skills I developed during that time have remained central to everything I do.

How did you get your first break?

My first break into the industry came with Ittiam Systems, after I returned to India. I was eager to contribute, and the company was working on vision-based ML problems for hardware-constrained environments, which aligned well with my academic background.

Getting a break in the industry after a PhD can definitely be challenging, but fortunately, it wasn’t too difficult in my case. I had already been freelancing during the final months of my PhD, which helped me stay connected to the industry and build a portfolio beyond academia.

I joined Ittiam Systems right after finishing the core work for my PhD, although I hadn’t yet defended my thesis at the time. Since no research work was pending, I was fully focused on my role at Ittiam while wrapping up the formal PhD defense process later.

But what really made the difference was my manager’s mentorship. He not only guided me technically but helped me understand the mindset required in a product-driven, real-world setup. That support and exposure built the foundation for everything I’ve pursued since in applied science and AI.

Ittiam is a product-based company focused on software — specifically, it’s a solutions provider for global leaders in video technology. They work on enabling next-generation experiences through a suite of intelligent video technologies and solutions. It was a great place to transition into industry, especially because it allowed me to apply and expand my technical skills in a fast-paced, product-driven environment.

What were some of the challenges you faced? How did you address them?

Challenge 1: Transitioning from Academia to Industry
In academia, you’re solving research problems with long timelines and academic rigor, whereas in the industry, it’s all about impact, speed, and iteration. Adapting to this mindset shift was one of the biggest challenges. I had to unlearn some habits and develop a solution-first approach instead of a paper-first approach. I invested time in understanding the domain, thinking from a user and product perspective, and learning how to prioritize outcomes over process.

Challenge 2: Technical Skill Gaps
In academia, I was primarily using MATLAB, but most real-world systems in the industry are built using Python, cloud tools, and scalable ML pipelines. I knew I had to catch up fast. I picked up Python, Docker, APIs, and other modern engineering tools through online resources, hands-on projects, and freelancing to bridge the gap.

Challenge 3: Industry Mindset
The way product decisions are made, how engineering teams operate, and how to balance trade-offs between performance and practicality—all of this was new. Thanks to my managers and mentors, I had many late-night discussions and planning sessions that helped me develop the mental model of a product engineer and leader, not just a scientist.

Where do you work now? What problems do you solve?

I currently lead the AI team at Parspec, a startup focused on Construction Tech.

Our core work revolves around:

  • Document processing
  • Recommendation engines
  • Search and retrieval systems using NLP, computer vision, and LLMs

The problems we solve include:

  • Making sense of technical product documents and spec sheets
  • Matching and recommending the right construction products from vendor catalogs

What skills are needed for this role? How did you acquire them?

  • NLP and Deep Learning
  • Document AI and OCR
  • Prompt engineering & LLM (Large Language Model) fine-tuning
  • System design, deployment, and monitoring
  • Team management and product planning

I acquired these through a combination of formal research experience, hands-on projects, mentorship, and most importantly, continuous learning and adapting to project needs.

What’s a typical day like?

  • Aligning team goals and deliverables
  • Planning and unblocking technical roadmaps
  • Diving into architecture or modeling details when needed
  • Mentoring and reviewing progress
  • Collaborating with the product and engineering teams
  • Experimenting with solutions for edge cases that don’t work out-of-the-box

What do you love about this job?

It’s challenging, fast-paced, and very research-heavy. Document processing is a domain where off-the-shelf solutions fail frequently, so there’s constant innovation and iteration. Being in a startup also means quick feedback loops, which I love—it’s the perfect mix of strategy, coding, research, and leadership.

How does your work benefit society? 

Construction is a massive industry that historically lacks automation and digital infrastructure. By applying AI:

  • We are streamlining tedious, manual workflows.
  • Reducing the time, cost, and errors in material selection, ordering, and documentation.
  • Empowering engineers and contractors to make smarter and faster decisions, which translates to efficiency and sustainability on real construction sites.

Ultimately, it’s about modernizing a legacy industry that touches infrastructure, housing, and cities—and doing it through tech.

Tell us an example of a specific memorable work you did that is very close to you!

My work at Walmart on fraud detection is something I hold very close.

It was a project that lacked initial support—not many stakeholders believed it could succeed. I took ownership, built the entire fraud detection module, and learned deployment and monitoring on-the-go. I also had the opportunity to work with multiple cross-functional teams across domains.

Eventually, it became one of the most impactful systems in the fraud prevention pipeline, and seeing something go from idea to scaled deployment was deeply satisfying. It taught me about ownership, resilience, and leadership under uncertainty.

Your advice to students based on your experience?

Don’t be afraid to start from behind—your background doesn’t define your future. Your effort, mindset, and adaptability do.

Learn how to learn. Technology changes fast, so the real skill is staying curious and flexible.

Ask questions, seek mentors, and don’t hesitate to admit what you don’t know.

Work on real problems—freelancing, open-source, hackathons—anything that pushes you out of theory and into messy, practical problem-solving.

And finally, remember: Failure is not the end. It’s just feedback. I failed mid-terms in 11th, but I topped in 12th. Life can turn around if you keep going.

Future Plans?

Looking ahead, my focus is on continuing to work at the intersection of AI research and real-world impact.

I want to:

  • Build AI systems that are not just intelligent, but reliable, explainable, and efficient, especially in industries like construction, healthcare, and finance where automation can create significant value.
  • Continue leading and scaling AI teams that combine research depth with product mindset. I enjoy solving open-ended problems and mentoring talent to build scalable, meaningful solutions.
  • I’m particularly excited about the future of LLMs and multi-modal AI—I see enormous potential in applying these to structured domains like document understanding and knowledge retrieval.