CIOInsider India Magazine
.png){kind=logo}
Kamalaakar Devaki
Founder & CEO
Technology is a powerful tool to evolve and impact lives, and the introduction of AI has been a gamechanger. AI hasn’t just impacted but also transformed every walk of our lives. Starting from process automation and diagnosis to decision making, AI is every where. While there has been an ongoing debate regarding AI eating up manual works and jobs, everyone would agree to the fact that instead of replacing jobs, AI is helping processes become more streamlined and automated. Various industries over the years have embraced AI to experience tremendous growth. The AI revolution in India has helped push many firms in the country to adopt AI as part of their offerings.
With a vision to work in the deep learning and the computer vision space, Kamalakar Devaki started Sandlogic in 2018. Today, with a market experience of around seven years, the firm has turned out to be one of the pioneering firms in the country to develop both small and large language models. The firm is also working edge chips that consumes just two watts of power while giving an astonishing 20 trillion operations per second. As one of the 13 organizations to get the grant from Indian government to build edge chips, Sandlogic has also been awarded with numerous awards, including Aegis Graham Bell Award and National Startup Award.
In an exclusive interview with CIO Insider, Kamalakar Devaki, Founder & CEO at Sandlogic Technologies, sheds further light on the firm’s flagship offerings, case studies, future roadmap, and much more:
Shed some light on your flagship edge computing solutions. What approach do you follow to maintain customer trust and satisfaction levels?
Our edge computing solutions can game changer. edge devices. On top of it, we have also built a layer that can help port AI workloads onto smaller devices. Apart from that, we also create models for specific use cases of the edge devices. While it serves a particular purpose, it also helps control the model and train it with customer-specific data for the use case before releasing it to the customer. Custom building models for the customers also come with additional data security and flexibility.
Instead of taking some model and force-feeding it, we at Sandlogic believe in developing custom models based on the requirement
Could you share a challenging case study reflecting the positive outcomes you drew and the pain points you addressed through one of your services?
In one of the instances, we helped a leading US-based warehouse operator with around 42 large warehouses streamline their operation. The firm was facing a huge challenge in managing the forklift operator. The forklift operators had to move in and out every time the team used to put a pallet. Previously, after placing the pallet, the team first used to remove their gloves, update their location, and then move out. The whole process was extremely time-consuming. So, we built an edge-based solution that was attached to the forklift. We also automated the location identification process for placing the pallet.
Even the process of picking up the right pallet by the forklift operator while moving out was automated, ensuring a quick turnaround time for the overall process. The whole process took place
around 2018-19 when putting these deep learning models into small devices was a huge challenge. From identifying the right camera to the model creation, we provided end-to-end support for the whole process.
In another case, we helped an Indian security organization build a camera for border security monitoring. This device could take input from multiple (around five) night vision cameras with a sensiti-vity range of over five km. Our job was to stitch the images coming from each of the cameras in real-time. It also had an intrusion detection feature.
To achieve that, we were dealing with around 1.8 billion pixels per second, making the whole project extremely challenging.
Describe the team behind building and deploying these solutions and the level of expertise they bring to the table.
At Sandlogic, instead of taking some model and force-feeding it, we believe in developing custom models based on the requirement. This includes identifying the requirement followed by creating the right algorithm. Based on that, datasets and methods are created to train the model. We have a highly accomplished team comprising PhD researchers, deep learning experts, embedded engineers, and FPGA engineers to solve this problem through a pure R&D approach. The team has a perfect mix of industry experience and domain expertise required for these projects.
Going forward, what is the future roadmap being planned for Sandlogic Technologies?
Currently, we are focused on completing the development work for the chip with a plan to launch the product by next June. Apart from that, we are also aiming to build a large language model. We are participating in the India AI mission, where we will be submitting our proposal to build India's large language model. We are glad to be a part of an industry that is currently gaining huge attention while driving growth.
