Rahul Tayal

A Hyperledger Fabric-based marketplace for seafood, which uses IoT to capture live a lobster's journey from sources like Australia to markets like China.

AI/ML-based tools enable easy integration of the lobster with its digital twin on a blockchain. The IoT data is then encrypted and synced with blockchain, so buyers know they are getting quality products and, at the same time, enjoy the lobster's whole journey like a story.

A real-time tele-ultrasound solution, enabled by 5G to realize the vision of an effective point-of-care system.

The solution has two main features. First, it has a medical collaboration system to connect the doctor or primary care professional and the patient for real-time diagnosis, anomaly detection, remote visualization, and consultation. Second, it has an AI-based doctor assistance and triaging system for expedited diagnostics, recommendations, and automated reports. Based on a configurable decision engine, the solution uses a microservices architecture with distributed edge cloud data aggregation and 5G multi-access edge computing to address real-time performance needs.

A mechanism that allows a resident to allocate consent to do a specific task to a service provider.

Previously, no system allowed UIDAI to do tasks such as filing 2022-2023 income tax returns or allowing an operator to update a resident's demographics stored in Aadhaar. As a solution, UIDAI envisaged the consent service, which I implemented. The service authenticates the resident and displays the list of privileges asked by the service provider. If the resident agrees, the system allocates a JSON Web Token to the operator for one task, allocating the approved privileges. It uses the OAuth2 protocol and is implemented in Spring Boot. Currently, the system has completed stage testing though not yet deployed.

Autonomous intelligent vehicles (AIVs) that use 5G to achieve complex, real-time computations to support fully autonomous inter-logistics operations, especially around a factory environment.

5G offers the promise of significant performance and cost benefits by offloading much of this computation from the AIV to the edge. The AIV is fitted with an inertial navigation system; hence there is dead reckoning or zero dependence on predefined tracks or pre-installed laser infrastructure for navigation. It uses camera-supplemented, simultaneous localization and mapping for correcting accumulated errors and increasing reliability. Its other features also include:

• Easy adaption to factory infrastructure. The solution can easily leverage existing cameras to bring down the overall cost of implementation.
• Real-time high precision pick up and drop and AI-enabled complex computations to dynamically detect and reorient the AIV for precise operations within 10 millimeters.
• Distributed management and control through 5G-enabled edge control algorithms, which are offloaded from the AIV to the edge server without degrading the solution's performance.

Developed an audio processing-based system with the following:

• Capability to have a conversation with humans;
• Convert audio to text;
• Using AI, make a probability distribution of the converted text and make corrections to the text if needed;
• Using NLP to detect the context and meaning of the sentence;
• Look up the user's current context and match it with the context and meaning of the current command from the user. Take any action required. Pull back the results.
• Create a response and convert it to text;
• Convert the text to audio and play it to the user.

Implemented the following using machine vision/image processing/object detection:

• Mine operations and safety monitoring using AI and object detection;
• Detecting road obstructions using AI and image processing;
• Creating blasting profiles based on terrain and requirement using AI;
• Vehicle safety monitoring and navigation using AI;
• Mine vehicle automation using 5G+MEC using AI and reinforcement learning.

My experience involves working on AI-based interactive systems that utilize speech-to-text and text-to-speech technology to create a full duplex communication platform. The product that I worked on was centered around automated DISA compliance-based auditing and solutions through a self-service voice-based system. Essentially, when a client called the system and provided authorization details for a remote system under audit, the system would securely log in and examine various services and assets to determine what needs to be tested. The client would then speak back with the services and assets that need to be tested, and further communications would occur between the client and the system until the system was clear on what needed to be done. Finally, the system would take further action and present a vocal report in a concise format while sending a detailed report via email.

The project that I worked on involved building an autonomous drone flight controller using inexpensive and easily accessible microcontrollers such as the STM32F103. The flight controller relied on several sensors including altitude sensors based on air pressure, an accelerometer and gyro, GPS, and a compass. To facilitate telemetry, we created a system using the Reyax rylr998. For manual operations and remote control, we utilized a PPM-enabled module with an IA6B receiver. To debug the system, we used an ST link V2 debugger attached to GDB. I was responsible for writing the entire flight controller, telemetry, and drone movement system, which had several advantages such as the ability to integrate with AI/ML models and adopt new sensors for various needs. Additionally, the system offered new flight modes and was very affordable due to the use of low-cost hardware.