CIO Insider

CIOInsider India Magazine


SandLogic Technologies: Simplifying AI to Solve Mission-Critical Problems of Organizations

   Kamalakar Devaki,     Founder

Autonomous systems, robots, smart homes/cities, medical devices, cleantech, energy efficiency, and much more growing or emerging application fields are all impacted by deep tech. Founded in 2018 by Kamalakar Devaki and his team, SandLogic (SL) Technologies is a full-stack enterprise AI company that provides LCNC platforms to develop deep learning applications to run on edge devices. Kamalakar and his team are born with a passion for deep-tech industry experts who came together to innovate tomorrow. And, build commercial-grade solutions using AI, ML, deep learning, computer vision, NLP, & edge AI to deliver the AI that powers enterprises. They are on a mission to enable enterprises to achieve the optimum force in their digital transformation & automation journey.

In conversation with Kamalakar Devaki, Founder, SandLogic Technologies

What made SandLogic Technologies begin its services based on the deep tech domain? How is your organization placed in the market?
When leading IT & product professionals came together to build on a common thematic of helping mid to large size companies in adopting AI much faster & simpler, that’s where SandLogic was bootstrapped. The team which started from a rented flat by Radhika Kanigiri, Ravi Kumar Rayana,Jesudas Fernandes & Kamalakar Devaki, has grown to 40 members today. And, operational in USA & India. Today, we are ultrafocused on deep learning & edge AI space. TXTR (OCR/ ICR)& (Edge AI) are the team’s current focus which spans Enterprise AI & Industrial AI respectively.

Furthermore, SandLogic was the first deep tech Startup to demonstrate a successful custom SoC based on Shakti

Vajra RISC-V processor combined with a deep learning accelerator that could run neural nets for image classification and object detection. We were the only team among 30 finalists of the Swadeshi microprocessor challenge organized by MeITY to demonstrate a custom SoC with a RISC-V and DLA on an FPGA board.

Today, we are ultra-focused on deep learning & edge AI space. TXTR (OCR/ICR) & EdgeMatrix. io (Edge AI) are the team’s current focus which spans Enterprise AI & Industrial AI respectively

Tell us about your portfolio of products and solutions that has made a
difference for your clients.

The key products which have helped customers in jump-starting their AI journey are CORE, - SL DLA, TXTR, and - CORE is for porting models from one AI framework to another, including target / hardware specific format. Whereas – SL DLA is for a complete end-to-end deep learning accelerator functional stack that can be integrated into different modes to realize AI chips or AI processing SoCs on FPGAs. Thirdly, TXTR is for a visual intelligence & NNLP-based intelligent document processing platform that aids in the automation of the business process/ workflows through capturing, classifying, verifying, & extracting insights. And lastly, AuVi. io is for data annotation & ML Ops platform for all AI project needs, right from data set preparation to model training & monitoring. All the products are built inhouse and will be soon out in the market as SaaS offerings.

Along with this, our firm has developed some unique solutions for our customers which includes road & pavement rating system, data & IP protection at endpoints (laptops & desktops), scan less pallet recognition & box-counting, scan less pallet sorting using edge AI attached for forklifts, and scan less catch weight using edge AI.

With us, customers can design, develop, train their AI models, and port/deploy

them onto small devices to cloud servers.

For such exceptional products and solutions one needs the latest technology, could you give a brief account of the latest technologies adopted by the company?
Under the AI and ML-related technologies and tools, SandLogic has adopted some latest trends. These technologies are NN development frameworks like TensorFlow, PyTorch, Keras, Caffe, ML.Net, Flutter, and TFLite, which provide neural nets in a different format. Then, the NN inter exchange format converter converts from one format to another and to a specific target-specific format. Thirdly, edge device and GPU specific SDKs and compute engines like TRT for Jetson, DLC for Snapdragon, VITIS AI for AMD Xilinx, OpenVINO for Intel, CUDA for NIVIDA GPUs, etc. And lastly, neural net deployment specific MLOps like serving frameworks like BentoML and torch serve.

Hence, with such extraordinary services, where is the company headed in the years to come?
Annual Edge AI processor shipments are forecast to reach 1.5 billion units by 2023, according to IDC. But, less than 0.1 percent of 1.5 billion are the Edge AI developers available in the market (who understand the hardware as well as AI software). SandLogic is filling that gap with its low code/no-code developer platform which helps organizations in automating the deep learning model porting on the edge life cycle.

Also, SandLogic is working on launching two one of its kind value-adds in the market which are ready to use soft DLA stack compatible with RISCV & ARM architecture and customized & optimized deep learning model zoo, as cloud API. With these major breakthroughs, the organizations can convert their non-AI hardware, much smarter & deploy AI models in hours rather weeks.

Current Issue
DriveBuddy AI: Leveraging Technology For Fleets & Logistics Management