India’s Addverb Brings AI and Edge Computing to Warehouse Robotics

Delivery timelines are shrinking, SKUs are proliferating and faster turnaround expectations have made warehouses one of the most intense pressure points in the supply chain. For Addverb, that pressure has translated into a deep engineering opportunity.

What began in 2016 as an idea to bring automation and standardization to warehousing has since evolved into a full-spectrum robotics and automation company based in Noida, Uttar Pradesh, India, with operations in the Middle East, Europe and Southeast Asia.

This growth is rooted in a consistent design philosophy that prioritizes in-house development across the full technology stack—from control systems to simulation and fleet management.

“From day one, we developed all the critical pieces, the robot controller, the navigation, the orchestration logic and even the visualization, in-house,” said Prateek Jain, co-founder and chief operating officer at Addverb.

Addverb designs and manufactures at its Noida-based Bot-Valley facility, a 2.5-acre campus housing its R&D lab, SMT line and manufacturing setup. The decision to own manufacturing was not just about control—it was a strategy to rapidly prototype, iterate and scale without dependency. The company is backed by Reliance Retail, which invested $132 million in 2021 and now owns a 55% stake. In 2023, Addverb expanded operations to a new facility in Greater Noida—Bot Verse—spread over 600,000 square feet, with a manufacturing capacity of 100,000 robots per year across various specifications and categories.

With a stack that now includes shuttles, mobile robots, control software, orchestration engines, vision systems and fleet managers, Addverb no longer identifies solely as a system integrator but as a made-in-India product development company.

“As of now, about 30 to 40% of the hardware and controls used in our products are localized. By the end of this year, we aim to increase this to 80 to 90%,” Jain said. “The software is fully developed in-house. Mechanically, we are already at 100% localization—we do not source any mechanical components from outside the country, except occasionally from China for high-volume requirements.”

The company has sold thousands of robots globally and counts major FMCG and e-commerce players—including Pepsico, UPS and Maersk—among its clients. Revenue numbers are undisclosed, but recent expansions in Singapore, Dubai and the Netherlands suggest growing traction.

Autonomy beyond automation

Initially, Addverb built fixed automation products like conveyors, and gradually moved to semi-automated systems like pallet shuttles. These evolved into fully automated storage systems like carton shuttles and mother-child shuttle systems. Today, the company has expanded into mobile robotics and even sorting robots based on barcode navigation.

“Our robots are powered by IPCs and include components like motor drivers, LiDARs and sensors,” Jain said. “If a component like a LiDAR is unavailable, we work around it—we’ve even developed custom logic cards and drivers ourselves.”

An image of Addverb's sorting robot Zippy. — Addverb’s sorting robot Zippy. (Source: Addverb)

“For mobile robots, we do all the mechanical design in-house. Our control team provides the required control systems. We also manufacture most of the boards we use in-house. We have our own assembly line,” he added. “However, for some components—like programmable logic controllers—we source directly from vendors like Siemens. For our own boards, we collaborate with companies like Arrow, DigiKey or directly with manufacturers such as STMicroelectronics. From parts to control system integration, everything is handled internally. We design, test prototypes and manufacture under one roof.”

Jain noted that most of the AI-related processing for mobile robots is done at the edge to minimize latency and enable faster operations.

“Latency matters,” he said. “You don’t want a robot waiting on a cloud server to decide whether to brake or turn.”

Navigation, manipulation and safety decisions are processed at the edge, while cloud computing is used for slower tasks like order analytics, image databases and long-term learning.

With a deliberate push toward physical AI, hybrid autonomy stacks and frictionless human-machine interfaces, Addverb is working toward a world where intelligent machines, whether humanoid, picking systems or mobile robots, can reason, adapt and collaborate in real time. Jain shared insight into the company’s current research and development initiatives.

Learning at the edge

One of the biggest challenges in logistics robotics is intelligent movement from point A to B amidst shifting shelves, human workers and unpredictable obstacles. Addverb’s Trakr 2.0 addresses this with what Jain described as a “layered autonomy stack.”

The Trakr 2.0 system integrates a 3D LiDAR, Intel RealSense camera and proprioceptive sensors to perceive its environment. All this feeds into an edge compute architecture based on Nvidia Jetson Orin and Intel CPUs, which process data on-device using deep reinforcement learning (DRL) models.

“We’re not sending data to the cloud and waiting for a decision,” Jain said. “Decisions about navigation, obstacle avoidance, and control are made in real time, right where the action is.”

Trakr 2.0 is designed to function like a warehouse worker with the muscle memory of a parkour athlete and the planning brain of a chess player. Its movement is governed by model predictive control and reinforced through DRL to optimize both energy use and agility.

An inventory system that thinks

On the other end of the warehouse, Addverb’s Horizontal Carousel System (HOCA) uses machine learning to predict SKU velocity, balance workload across picking stations and forecast demand.

“HOCA doesn’t just follow commands from a warehouse management system—it feeds insights back into it,” Jain said.

It also blends computer vision and deep learning to validate picks and reduce errors—without requiring workers to constantly scan barcodes or double-check screens. The goal is not to replace humans, but to augment them with systems smart enough to anticipate and assist.

Intuitive robotics

If Trakr 2.0 and HOCA sound advanced, Addverb’s human-machine interface (HMI) project, Brisk, flips the script. It enables gesture-based control in warehouse settings—no touch screens or buttons required.

“We realized that many users in warehouses aren’t tech-savvy—and they shouldn’t have to be,” Jain said. “With vision-based gesture control, even someone who’s never worked with a robot can walk up and start using one.”

The system uses high-resolution cameras and lightweight ML models to translate gestures into machine instructions. Future versions will include voice controls, AR overlays, and projection-based guidance to create a more inclusive, low-friction interface.

Leaping toward humanoid bots

While Brisk simplifies interaction, Addverb’s humanoid robot ventures into far more complex territory. Still in early development, the bot exemplifies the idea of physical AI—where intelligence is embodied in movement, balance and adaptive reaction.

It comes with a hierarchical control system, which includes a low-level motor control handled by STM32 microcontrollers and high-level cognition by Jetson Orin and Intel Tiger Lake i7 processors. It uses sim-to-real transfer learning, domain randomization and feedback loops to perform manipulation, inspection and work in hazardous environments.

“We’re not just programming fixed behaviors,” Jain said. “The robot learns from its environment and adjusts over time. That’s physical AI in action.”

This is not about bipedal robots walking around warehouses—it is about building systems that can handle repetitive, unstructured tasks that humans struggle to stay engaged with.

With so many advanced developments underway, Addverb is currently focused on increasing investments in simulation tools, sensors and compute hardware, expanding its R&D team (now over 400 engineers) and strengthening its global sales and marketing engine.

The company is actively seeking channel partners, especially system integrators with local warehousing expertise, electronics vendors for high-efficiency power systems, edge computing and sensors, as well as academic collaborators for DRL and HMI-focused research.

“The ecosystem has a big role to play,” Jain said. “We’re looking for collaborators who can bring domain knowledge, supply chain strength or research depth.”

Whether it is creating robots that sense with instinct, pick with precision or listen with understanding, Jain emphasized the company’s core intent: to build smarter, faster and more intuitive automation powered by full-stack engineering and relentless curiosity.

There is a quiet confidence in the way Jain discussed these technologies: less hype, more focus on practical implementation. Addverb’s goal is not to flood warehouses with robots—it is to revamp them completely.

“Motion without intelligence is just noise,” he said. “We’re designing machines that evolve: physically, cognitively and contextually.”