About
Aryaveer Singh Parihar is a Visiting Researcher at the Cognitive Interface Lab and a Computer Science and Engineering undergraduate at Shri Mata Vaishno Devi University, with experience in edge AI and embedded systems. His research focuses on hardware-software co-design for efficient AI, enabling wearables and other low-power devices to respond to people in real time without depending on the cloud. He also works on reducing the carbon footprint of AI, studying how to run large language model inference with lower energy and emissions.
Research Interests
- Hardware-Software Co-Design
- Embodied AI
- Human-Computer Interaction
- Model Compression for Edge AI
- Sustainable and Carbon-Aware Computing
Projects
Carbon-Aware Scheduling for Large Language Model Inference
Ongoing · University of Ljubljana
A scheduler that decides where a query runs (on the user’s device, an edge server, or a data centre) and with which model size and quantisation. It aims to cut carbon emissions while meeting latency and answer-quality targets.
Nida-e-Watan: Offline Edge Inference Cluster
A 12-node Raspberry Pi 4 cluster running compressed multilingual language models (1.1 GB to 260 MB, keeping 91% of task accuracy) for villages with limited connectivity.
Behaviour-Based Detection of USB Charging Threats
A detection pipeline and telemetry simulator for malicious USB charging sessions. It is evaluated under deliberately hard conditions, and the scorer is exported to 27 KB of dependency-free C.
Eco-Smart Wearable Biosensing Patch
A cloud-free biosensing patch on an Arduino Nano that runs a time-series classifier on the microcontroller itself.
Project code is not publicly available. The carbon-aware scheduling work is ongoing and unpublished; some project data is private or subject to licensing restrictions.
