RSRahul Sundar

Rahul Sundar

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Rahul Sundar has a total of 11 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    InnoGuideGPT: Integrating conversational interface and command interpretation for navigation robots

    by Rahul Sundar, Shreyash Gadgil, Tankala Satya Sai, Sathi Sai Krishna Reddy, Gautam B, Ishita Mittal, Jyotsna Sree Guduguntla, Shanmukesh Pujala

    Integrating natural language understanding, voice command interpretation and natural language generation for realtime inference is a challenging problem. However, developing a proof of concept is now possible in just a few lines of code which was otherwise unimaginable a few years ago. Thanks to the democratization of LLMs and the availability of high-quality pre-trained models through APIs. It is now possible to quickly build effective use cases by just integrating multiple models without even having to pre-train/fine-tune the models on custom data. This is due to their zero-shot learning ability. Although, there are many recent works in this regard oriented towards software [1, 5], edge implementations and their applications to Robotics are yet to be explored in their entirety [6]. Recently, Koubaa [2] proposed RoboGPT, where a ChatGPT was prompt-tuned for command interpretation and subsequently determining the robot’s actions. In this work, we explore the possibility of integrating a real-time conversational voice interface within a navigation robot. This is achieved using an underlying LLM interface to respond to user queries based on a prior context and additionally interpret voice commands for the robot to navigate. As a proof of concept, we evaluate the effectiveness of the planned workflow using a robot simulation. For simulating a Robot’s environment, there exist various tools but in this work, Gazebo [3] has been adopted to simulate a differential drive robot moving in an indoor environment and Rviz is used for visualization. We utilized the TurtleBot3 [4] software packages to implement motion planning and navigation algorithms (see figure 2). For the speech recognition, we use OpenAI’s Whisper API, while for text to speech we use the freely available google text-to-speech python package. We then use OpenAI’s GPT-3.5-turbo API within the Langchain framework for building a context aware conversational interface and an additional command interpretation module under the hood to guide the navigation bot. It is remarkable how there is a significant reduction in development time for proof of concepts through the availability of high-quality models in addition to high-quality outcomes. Once the workflow is validated on Gazebo, the workflow is then implemented onto a NVIDIA Jetson Nano which will be used to send and process the cloud based LLM/speech API requests and responses (see figure 3.


  • Open Access

    Keyword Augmented Retrieval: Novel framework for Information Retrieval integrated with speech interface

    by Anupam Purwar, Rahul Sundar

    Retrieving answers in a quick and low cost manner without hallucinations from a combination of structured and unstructured data using Language models is a major hurdle. This is what prevents employment of Language models in knowledge retrieval automation. This becomes accentuated when one wants to integrate a speech interface on top of a text based knowledge retrieval system. Besides, for commercial search and chat-bot applications, complete reliance on commercial large language models (LLMs) like GPT 3.5 etc. can be very costly. In the present study, the authors have addressed the aforementioned problem by first developing a keyword based search framework which augments discovery of the context from the document to be provided to the LLM. The keywords in turn are generated by a relatively smaller LLM and cached for comparison with keywords generated by the same smaller LLM against the query raised. This significantly reduces time and cost to find the context within documents. Once the context is set, a larger LLM uses that to provide answers based on a prompt tailored for Q&A. This research work demonstrates that use of keywords in context identification reduces the overall inference time and cost of information retrieval. Given this reduction in inference time and cost with the keyword augmented retrieval framework, a speech based interface for user input and response readout was integrated. This allowed a seamless interaction with the language model.