Recent Research Projects
Energy-aware localization for passive backscatter tags
February 2024 - Present
Using passive backscatter tags for localization is a novel idea, but the lack of an active transceiver makes inter-tag distance measurements ambiguous. This project enabled machine-learning-based unambiguous ranging between passive backscatter tags with a median error of approximately 6.9cm for distances up to 2m. We also achieved a median localization error of less than 2.5 cm with only 12 anchors and proposed a heuristic to measure only high-quality links for localization, resulting in power savings of up to 51% while minimally affecting localization accuracy.
Benchmarking vision tasks on edge and cloud
February 2024 - Present
Real time learning based vision applications involve diverse tasks, ranging from simple object detection to self-driving. Deciding on deployment and execution, whether distributed, monolithic, edge-based, cloud-based, or hybrid, is often not trivial. To aid inecision making, this work strives to provide comprehensive benchmarking for several of the most relevant real-time tasks with multiple state-of-the-art methods
Collaborative Backscatter in passive backscatter tags
January 2023 - July 2024
Passive backscatter is a promising technology for scalable IoT, but it suffers from shorter communication ranges compared to active RF. This work developed methods to allow helper tags to configure their almost-optimal backscatter phase in polynomial time from an exponential search space. Experiments showed a median gain of about 20% with static help and a 3-4x gain with active help from helper tags. This work was accepted at the 2025 IEEE International Conference on RFID.
Distributed Modular Video Analytics on the edge
August 2022 - July 2024
This project improved the accuracy of modular multi-model Video Analytics (VA) pipelines on cheap, low-compute, heterogeneous edge devices to minimize reliance on the cloud. We proposed an application-aware module placement and replication heuristic to maximize throughput and application accuracy. Additionally, we developed an adaptive model selection algorithm for DNN-based VA modules to maximize accuracy in response to varying load conditions. When compared against several baselines in our evaluation, we achieved throughput and accuracy gains of at least 51% and 28%. This work was accepted at the Symposium on Edge Computing 2024 (SEC 2024).
Work Experience
Research Intern | NEC Labs America
June 2025 - August 2025
- Proposed an agentic-AI pipeline to improve LLMs' tool selection and calling using a hybrid approach.
- Evaluated the method on Nestful and BFCLv3 benchmarks, achieving 27% and 3% improvements, respectively, over direct LLM use (GPT-4o).
Research Intern | Tata Research Development and Design Centre
June 2021 - July 2021
- Evaluated the security of numerous sign-up interface designs and gathered data from about 100 web and mobile applications of various categories.
- Created an evaluation platform with combinations of selected top interface design choices to perform A/B testing.
Publications
- [Poster]Manavjeet Singh, Yang Xie, Abeer Ahmad, Milutin Stanaćević, Samir R Das, Petar M Djurić, "Tag-to-Tag Range Estimation For Passive Backscatter Tags," International Conference on COMmunication Systems & NETworkS (COMSNETS) 2026, Bangaluru, India.
- [PDF]Manavjeet Singh, Kunal Rao, Giuseppe Coviello, Srimat Chakradhar, "TacTool: Tactical Tool Usage in Agentic AI Systems," IEEE International Conference on Agentic AI (ICA) 2025, Wuhan, China.
- [PDF]Abeer Ahmad, Manavjeet Singh, Yang Xie, Xiao Sha, Milutin Stanaćević, Samir R Das, Petar M Djurić, "Improving Communication Performance of Passive Backscattering Tags Using Collaborative Backscatter," 2025 IEEE International Conference on RFID (RFID), Atlanta, GA, USA.
- [PDF]Manavjeet Singh, Sri Pramodh Rachuri, Bryan Bo Cao, Abhinav Sharma, Venkata Bhumireddy, Francesco Bronzino, Samir R Das, Anshul Gandhi, Shubham Jain, "OVIDA: Orchestrator for Video Analytics on Disaggregated Architecture," IEEE/ACM Symposium on Edge Computing (SEC) 2024, Rome, Italy.
- Yang Xie, Yang Li, Manavjeet Singh*, Samir R Das, Petar M Djurić, Milutin Stanaćević, "Robust and energy-efficient channel estimation in RF backscatter tag-to-tag network," IEEE Journal of Radio Frequency Identification 2025.
- Anshak Goel*, Deeptorshi Mondal*, Manavjeet Singh*, Sahil Goyal, Navneet Aggarwal, Jian Xu, Mukulika Maity, Arani Bhattacharya, "FlexDisplay: An optimized smartphone display framework to conserve battery power," ACM J. Comput. Sustain. Soc., Jul. 2025.
- Bryan Bo Cao, Abhinav Sharma, Manavjeet Singh, et al., "Representation Similarity: A Better Guidance of DNN Layer Sharing for Edge Computing without Training," ACM International Conference on Mobile Computing and Networking, S3 Workshop 2024, Washington, DC, USA.
- Anshak Goel, Deeptorshi Mondal, Manavjeet Singh, Sahil Goyal, Navneet Aggarwal, Jian Xu, Mukulika Maity, Arani Bhattacharya, "FlexDisplay: A Flexible Display Framework To Conserve Smartphone Battery Power," IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops) 2024, Biarritz, France.