Interview with Rahul Saxena, co-founder and CTO at AiDash

Any team should always have a mix of people with varying skills and exposure to run the business without compromises.

Sustainability has become a buzzword in business due to growing environmental concerns. An increasing number of companies have committed to the cause, setting lofty goals for themselves. But how do we measure the impact that they have had before we go about setting targets? Bengaluru-based startup AiDash believes it can help answer this important question.

Founded in 2019, AiDash is an AI-first vertical SaaS company enabling satellite- and AI-powered operations, maintenance, and sustainability for industries with geographically distributed assets. AiDash uses high-resolution multispectral and SAR data from the world’s leading satellite constellations that are fed into its proprietary AI models to make timely predictions at scale. These AI models empower AiDash’s full-stack applications that transform Operations & Maintenance for utility, energy, transportation, water and wastewater, mining, and construction companies.

THE BELAMY

Sign up for your weekly dose of what's up in emerging technology.

AiDash has a fully functional AI infrastructure that was built ground-up to cater to the specific needs of a business. We spoke with the co-founder and CTO of the company, Rahul Saxena, about the business end of setting up tech infrastructure and how AiDash overcame the challenges that come with it. 

AIM: The common belief is that using AI infrastructure costs a fortune. Is this a myth? What are the ways through which AiDash reduced costs to develop AI solutions?

Rahul Saxena: Building an AI infrastructure becomes costly when the groundwork is not well-thought-through and implemented as the framework must handle and train huge amounts of data, account for computation, storage, and data scaling. Having said that, setting up robust frameworks that reduce operational costs over time and align to the business needs can be an optimal solution.

At AiDash, we overcame these challenges by implementing the following strategies:

–   Adoption of cloud technology

Our entire AI/ML stack is built on the cloud, which is used very optimally for storage and computation. This also provides us with reliable and fast networks, scalability options, and secure delivery models.

–   Utilising open-source technologies

We analysed and adopted the best open-source technologies that cater to our needs and do not incur additional expenses. While many AI/ML open-source tools are gaining traction due to their numerous collaborative benefits, AI and big data companies are also starting to provide commercial support and add-ons to open-source AI tools.

         –            Attracting skilled talent

We have with us highly skilled data scientists and ML engineers who help us build accurate and robust algorithms. They know the ins and outs of building a successful data science product. And this proficiency leaves very little room for errors and increases the success rate of the products.

In short, businesses can develop pathbreaking AI solutions at an optimal cost; if the groundwork is done well, the best technologies that provide ROI, in the long run, are adopted, and the right people are hired.

AIM: How did AiDash select its ML model, and did it require consultation from experts?

Rahul Saxena: Selecting an ML model is a gradual process that requires a deep understanding of the requirements and problems and continuous experimentation to achieve the right model for a business. We need to factor in different parameters like data, accuracy, precision, classification, prediction, and many more while selecting an ML model, which requires a lot of time and evaluation.          

At AiDash, we did not require any external consultation because our AI and data science functions are led by Nitin Das, co-founder and Chief AI Officer, who is an AI-ML guru. He also has deep expertise in neural networks and satellite analytics and has solid experience in driving AI/ML initiatives and development from scratch. Our data science team is highly competitive and consists of expert data scientists and ML engineers who are capable of driving AI model developments from research to deployment.

AIM: Should startups opt for custom AI solutions for their software or third-party AI software? Does the difference in costs factor into the decision? How did you go about making the decision?

Rahul Saxena: We have experienced that opting for a custom AI solution is a better approach, though it mainly depends on the target use cases and complexity of the product. Building a custom solution includes building the custom algorithms and proprietary APIs, which involves upfront costs and setting up infrastructure. But this proved beneficial in the long run, paving the way for data customisation, reduced operational/support costs, and increased the quality of predictions for specialised data.

We have built all our AiDash products using custom AI solutions because we cater to a niche industry and have very specific use cases. Our bespoke AI products have the best prediction capabilities, can handle huge volumes of data, and are leading in the industry owing to the robustness and versatility of our AI models.

Because our product line is growing constantly, finding third-party software that encompasses all our current and growing needs did not seem like a feasible solution. And third-party AI programs can prove costly in the long run as the business and requirements expand.

AIM: How important are APIs in core ML and AI businesses?

Rahul Saxena: APIs play a key role in AI and ML development as they open a whole lot of possibilities to an existing solution, such as reusability and scalability.

Reusability is a key factor for optimisation in AI/ML model development. A model is not optimal if it does not go into production and is not reusable across platforms or products. To make AI and ML models reusable and scalable, we require APIs and automation. It also provides additional security to your code.

AIM: How difficult is it to estimate the various costs involved in AI? How was your experience in this regard?

Rahul Saxena: A lot of cost components are involved in setting up a scalable AI infrastructure. The primary one is the cloud technology that is used for computations, storage, and networking. Other factors are around developing the custom algorithms or adopting third-party software, which can add up to the operational costs.

When setting up the infrastructure at AiDash, we involved relevant AI/ML and DevOps experts during the planning and implementation phases to ensure all aspects were analysed and accounted for. We believe that analysing our requirements thoroughly with a futuristic vision and choosing the best approach helped us set up the infrastructure that aligns with our strategic goals.

AIM: What are some of the valuable lessons that you learned during this process?

Rahul Saxena: Setting up an AI infrastructure is a gradual process and requires an in-depth understanding of key requirements, models, data, computation, and storage.

  • We learned that starting with manual storage and collecting computation requirements for each model profiling could be used to automate the process to set up a framework that can be scaled and keeps evolving as we grow rather than targeting the complete stack.
  • While our initial AI/ML software stack was tightly coupled with a cloud provider, we later realised about additional critical capabilities offered by other cloud providers and then moved on to developing a cloud-agnostic stack. This has enabled us to have a software stack that can be moved across different cloud providers easily.
  • Initially, we were using the cloud-based GPU infrastructure but later realised that it would be rather costly as our requirements increased. So, we have now adopted a combination of in-house and cloud-based GPU infrastructure.

From all our learnings, today, we have built systems that can estimate the precise storage and computation requirements for each model we develop.

AIM: How important is figuring out team composition for an AI startup? What is the first step to building your team?

Rahul Saxena: Like any startup, an AI startup also requires all-rounders who can play various roles and have great analytical and adaptational skills. When starting the journey with an AI team, you require engineers with a vision who have expertise in researching and developing AI models along with software development skills to visualise and implement AI/ML solutions. But as you foray into the deep and are beginning to expand, you need to bring in expert data scientists, ML engineers, and DevOps engineers with relevant exposure and niche skills on your team. Any team should always have a mix of people with varying skills and exposure to run the business without compromises.

At AiDash, we believe in hiring the best and have a highly talented team of data scientists, ML engineers, and DevOps personnel who contribute to software development, data engineering, API development, data modelling, and so on.

More Great AIM Stories

Poulomi Chatterjee
Poulomi is a Technology Journalist with Analytics India Magazine. Her fascination with tech and eagerness to dive into new areas led her to the dynamic world of AI and data analytics.

Our Upcoming Events

Masterclass, Virtual
How to achieve real-time AI inference on your CPU
7th Jul

Masterclass, Virtual
How to power applications for the data-driven economy
20th Jul

Conference, in-person (Bangalore)
Cypher 2022
21-23rd Sep

Conference, Virtual
Deep Learning DevCon 2022
29th Oct

3 Ways to Join our Community

Discord Server

Stay Connected with a larger ecosystem of data science and ML Professionals

Telegram Channel

Discover special offers, top stories, upcoming events, and more.

Subscribe to our newsletter

Get the latest updates from AIM
MOST POPULAR

What can SEBI learn from casinos?

It is said that casino AI technology comes with superior risk management systems compared to traditional data analytics that regulators are currently using.

Will Tesla Make (it) in India?

Tesla has struggled with optimising their production because Musk has been intent on manufacturing all the car’s parts independent of other suppliers since 2017.