arXiv subjects

Prasoon Patidar

Publications and source records attributed to Prasoon Patidar.

4 recordsLinked to original sources

arXiv2026

OrganicHAR: Towards Activity Discovery in Organic Settings for Privacy Preserving Sensors Using Efficient Video Analysis

Deploying human activity recognition (HAR) at home is still rare because sensor signals vary wildly across houses, people, and time, essentially requiring in-situ data collection and training. Prior approaches use cameras to generate training labels for privacy-preserving sensors (LiDAR, RADAR, Thermal), but this forces sensors to detect predefined activities that cameras can see yet the sensors themselves cannot reliably distinguish. In this work, we introduce OrganicHAR, an activity discovery framework that inverts this relationship by placing sensor capabilities at the center of activity discovery. Our approach identifies naturally occurring signal patterns using privacy-preserving sensors, leverages Vision Language Models (VLMs) only during these key moments for scene understanding, and discovers discrete activity labels at granularities that these sensors can reliably detect. Our evaluation with 12 participants demonstrates OrganicHAR's effectiveness: it achieves 79% accuracy for coarse (4-5) activities using only basic ambient sensors (radar, lidar, thermal arrays), and 73% accuracy for fine-grained (8-9) activities when a wearable IMU, depth, and pose sensor are added. OrganicHAR maintains 77% accuracy on average across configurations while discovering 4-8 categories per user (15 across all users) tailored to each environment and sensor capabilities. By triggering video processing only at key moments identified by local sensors, we reduce queries to VLM by 90%, enabling practical and privacy-preserving activity recognition in natural settings.

cs.HC↗

arXiv2025

Orchestration for Domain-specific Edge-Cloud Language Models

The remarkable performance of Large Language Models (LLMs) has inspired many applications, which often necessitate edge-cloud collaboration due to connectivity, privacy, and cost considerations. Traditional methods primarily focus on selecting the best LLM model for optimizing performance, while neglecting the critical interplay between the components of the LLM serving pipeline (context retrieval, query preprocessing, etc.) or the changing latency and cost constraints. We introduce ECO-LLM (Edge-Cloud Orchestrator for LLMs), a novel system that reframes this problem as a joint optimization challenge and solves it by systematically exploring component configurations and dynamically selecting optimal strategies at the query level. ECO-LLM consists of two components: (1) the ECO-LLM Emulator, which efficiently explores the vast configuration space utilizing query clustering and pareto-optimal path selection, gathering domain-specific performance metrics without exhaustive evaluation; and (2) the ECO-LLM Runtime, which leverages these metrics to dynamically select optimal resolution strategies for user queries while meeting user-defined Service Level Objectives (SLOs). We evaluate ECO-LLM on a smart home and a smart car assistant scenarios. With an exhaustive exploration of all possible configurations for seen queries, ECO-LLM outperforms cloud-based models like GPT-4o in terms of accuracy (90% vs. 74% on average) while reducing costs by 90% and latency by 55%, demonstrating the value of its joint optimization at the query level. In practical deployment for previously unseen queries, ECO-LLM selects configurations that reduce costs by 62% or improve response times by 62% on average compared to state-of-the-art model routing approaches, while maintaining higher accuracy and consistently adhering to specified latency and cost constraints.

cs.DB↗

arXiv2023

Multi-Purchase Behavior: Modeling, Estimation and Optimization

We study the problem of modeling purchase of multiple products and utilizing it to display optimized recommendations for online retailers and e-commerce platforms. We present a parsimonious multi-purchase family of choice models called the Bundle-MVL-K family, and develop a binary search based iterative strategy that efficiently computes optimized recommendations for this model. We establish the hardness of computing optimal recommendation sets, and derive several structural properties of the optimal solution that aid in speeding up computation. This is one of the first attempts at operationalizing multi-purchase class of choice models. We show one of the first quantitative links between modeling multiple purchase behavior and revenue gains. The efficacy of our modeling and optimization techniques compared to competing solutions is shown using several real world datasets on multiple metrics such as model fitness, expected revenue gains and run-time reductions. For example, the expected revenue benefit of taking multiple purchases into account is observed to be $\sim5\%$ in relative terms for the Ta Feng and UCI shopping datasets, when compared to the MNL model for instances with $\sim 1500$ products. Additionally, across $6$ real world datasets, the test log-likelihood fits of our models are on average $17\%$ better in relative terms. Our work contributes to the study multi-purchase decisions, analyzing consumer demand and the retailers optimization problem. The simplicity of our models and the iterative nature of our optimization technique allows practitioners meet stringent computational constraints while increasing their revenues in practical recommendation applications at scale, especially in e-commerce platforms and other marketplaces.

cs.IR↗

arXiv2022

Self-Serviced IoT: Practical and Private IoT Computation Offloading with Full User Control

The rapid increase in the adoption of Internet-of-Things (IoT) devices raises critical privacy concerns as these devices can access a variety of sensitive data. The current status quo of relying on manufacturers' cloud services to process this data is especially problematic since users cede control once their data leaves their home. Multiple recent incidents further call into question if vendors can indeed be trusted with users' data. At the same time, users desire compelling features supported by IoT devices and ML-based cloud inferences which compels them to subscribe to manufacturer-managed cloud services. An alternative to use a local in-home hub requires substantial hardware investment, management, and scalability limitations. This paper proposes Self-Serviced IoT (SSIoT), a clean-slate approach of using a hybrid hub-cloud setup to enable privacy-aware computation offload for IoT applications. Uniquely, SSIoT enables opportunistic computation offload to public cloud providers while still ensuring that the end-user retains complete end-to-end control of their private data reducing the trust required from public cloud providers. We show that SSIoT can leverage emerging function-as-a-service computation (e.g. AWS Lambda) to make these offloads cost-efficient, scalable and high performance as long as key limitations of being stateless, limited resources, and security isolation can be addressed. We build an end-to-end prototype of SSIoT and evaluate it using several micro-benchmarks and example applications representing real-world IoT use cases. Our results show that SSIoT is highly scalable, as compared to local-only approaches which struggle with as little as 2-4 apps in parallel. We also show that SSIoT is cost-efficient (operating a smart doorbell for $10 a year) at the cost of minimal additional latency as compared to a local-only hub, even with a hardware ML accelerator.

cs.CR↗