This study investigates Shiksha Copilot, an AI-assisted lesson planning tool deployed in government schools across Karnataka, India. The system combined LLMs and human expertise through a structured process in which English and Kannada lesson plans were co-created by curators and AI; teachers then further customized these curated plans for their classrooms using their own expertise alongside AI support. Drawing on a large-scale mixed-methods study involving 1,043 teachers and 23 curators, we examine how educators collaborate with AI to generate context-sensitive lesson plans, assess the quality of AI-generated content, and analyze shifts in teaching practices within multilingual, low-resource environments. Our findings show that teachers used Shiksha Copilot both to meet administrative documentation needs and to support their teaching. The tool eased bureaucratic workload, reduced lesson planning time, and lowered teaching-related stress, while promoting a shift toward activity-based pedagogy. However, systemic challenges such as staffing shortages and administrative demands constrained broader pedagogical change. We frame these findings through the lenses of teacher-AI collaboration and communities of practice to examine the effective integration of AI tools in teaching. Finally, we propose design directions for future teacher-centered EdTech, particularly in multilingual and Global South contexts.

Large Language Models (LLMs) have significantly impacted the field of Math Word Problems (MWPs), transforming how these problems are approached and solved, particularly in educational contexts.However, existing evaluations often focus on final accuracy, neglecting the critical aspect of reasoning capabilities.This work addresses that gap by evaluating LLMs' abilities to detect and correct reasoning mistakes.We present a novel dataset, MWP-MISTAKE, containing MWPs with both correct and incorrect reasoning steps generated through rule-based methods and smaller language models.Our comprehensive benchmarking of state-of-the-art models such as GPT-4o and GPT4 uncovers important insights into their strengths and limitations.While GPT-4o excels in mistake detection and rectification, gaps remain, particularly in handling complex datasets and novel problems.Additionally, we identify concerns with data contamination and memorization, which affect LLM reliability in real-world applications.While OpenAI' O1 model demonstrates 90% accuracy in reasoning and final answers on complex tasks, it remains weak in mistake detection.Our findings highlight the need for improved reasoning evaluations and suggest ways to enhance LLM generalization and robustness in math problemsolving.Question: What is the product of the coordinates of the midpoint of a line segment with endpoints at (1,1) and (-7,5)? Reasoning Step:The midpoint of a line segment is located at the point (h,k), where h = (x1 + x2)/2 and k = (y1 + y2)/2.In this case, the endpoints of the line segment are (1,1) and (-7,5).So, we can find the midpoint by calculating h = (1 + (-7))/2 = 6 and k = (1 + 5)/2 = 3.Therefore, the product of the coordinates of the midpoint is hk = 6 * 3 = 18.Reasoning chain Correct (Yes/No): No Corrected reasoning chain or NA: The midpoint of a line segment is located at the point (h,k), where h = (x1 + x2 )/ 2 and k = (y1 + y2)/2.In this case, the endpoints of the line segment are (1,1) and (-7,5).So, we can find the midpoint by calculating: h = (1 + (-7))/2 = -6/2 = -3 k = (1 + 5)/2 = 6/2 = 3 Therefore, the product of the coordinates of the midpoint is: hk = -3 * 3 = -9 Final answer (just the number): -9 Reasoning chain Correct (Yes/No): Yes Correct Reasoning Chain or NA: NA Final answer (just the number): 18 Model Input Open AI Models GPT-4o GPT 3.5 Turbo You are provided with a mathematical question and a step-by-step solution along with it.The solution might have some mistakes.Identify if the solution is correct or incorrect.If the solution is correct, output the final answer with the help of the solution provided.If the solution is incorrect, correct the existing solution and determine the final answer with the help of the corrected solution.

Large language models (LLMs) have revolutionized various domains but still struggle with non-Latin scripts and low-resource languages. This paper addresses the critical challenge of improving multilingual performance without extensive fine-tuning. We introduce a novel dynamic learning approach that optimizes prompt strategy, embedding model, and LLM per query at runtime. By adapting configurations dynamically, our method achieves significant improvements over static, best and random baselines. It operates efficiently in both offline and online settings, generalizing seamlessly across new languages and datasets. Leveraging Retrieval-Augmented Generation (RAG) with state-of-the-art multilingual embeddings, we achieve superior task performance across diverse linguistic contexts. Through systematic investigation and evaluation across18 diverse languages using popular question-answering (QA) datasets we show our approach results in 10-15% improvements in multilingual performance over pre-trained models and 4x gains compared to fine-tuned, language-specific models.

Hengyi Wang, Haizhou Shi, Shiwei Tan, Weiyi Qin, Wenyuan Wang, Tunyu Zhang, Akshay Nambi, Tanuja Ganu, Hao Wang. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Passive Infra-Red (PIR) sensors are ubiquitous and have applications ranging from automatic lighting and heating control in smart buildings, towel dispensers in washrooms, security alarms (for intrusion detection) to human detection robots (for search and rescue). Unfortunately, PIR sensors are prone to failures during deployment due to reasons such as environmental damage, incorrect installation and component degradation among others that can lead to incorrect or faulty data. Currently, such failures are typically detected using either : (a) heavily engineered data-driven, statistical approaches that can have high false positive rates due to unseen data patterns or (b) expensive, unscalable methods that use additional hardware such as video cameras or a golden reference sensor. In this work, we first create a taxonomy for the most common PIR sensor failures and analyze these failures from the perspective of sensor physics. We then present PIRMedic— a physics-driven, edge-based approach to detect and diagnose the failures in a PIR sensor using an intrinsic hardware signal viz., the analog output from the pyroelectric element in the sensor. Using this hardware signal in conjunction with frequency analysis and supervised machine learning methods, we obtain a high accuracy of 98-99% in failure detection and diagnosis. We evaluate our methods using multiple real-world deployments , in four distinct locations, in different environment and usage conditions.

Spatial representation learning (SRL) aims at learning general-purpose neural network representations from various types of spatial data (e.g., points, polylines, polygons, networks, images, etc.) in their native formats. Learning good spatial representations is a fundamental problem for various downstream applications such as species distribution modeling, weather forecasting, trajectory generation, geographic question answering, etc. Even though SRL has become the foundation of almost all geospatial artificial intelligence (GeoAI) research, we have not yet seen significant efforts to develop an extensive deep learning framework and benchmark to support SRL model development and evaluation. To fill this gap, we propose TorchSpatial, a learning framework and benchmark for location (point) encoding,which is one of the most fundamental data types of spatial representation learning. TorchSpatial contains three key components: 1) a unified location encoding framework that consolidates 15 commonly recognized location encoders, ensuring scalability and reproducibility of the implementations; 2) the LocBench benchmark tasks encompassing 7 geo-aware image classification and 10 geo-aware imageregression datasets; 3) a comprehensive suite of evaluation metrics to quantify geo-aware models’ overall performance as well as their geographic bias, with a novel Geo-Bias Score metric. Finally, we provide a detailed analysis and insights into the model performance and geographic bias of different location encoders. We believe TorchSpatial will foster future advancement of spatial representationlearning and spatial fairness in GeoAI research. The TorchSpatial model framework and LocBench benchmark are available at https://github.com/seai-lab/TorchSpatial, and the Geo-Bias Score evaluation framework is available at https://github.com/seai-lab/PyGBS.

Kabir Ahuja, Harshita Diddee, Rishav Hada, Millicent Ochieng, Krithika Ramesh, Prachi Jain, Akshay Nambi, Tanuja Ganu, Sameer Segal, Mohamed Ahmed, Kalika Bali, Sunayana Sitaram. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023.

Real-time perception requires planned resource utilization. Computational planning in real-time perception is governed by two considerations -- accuracy and latency. There exist run-time decisions (e.g. choice of input resolution) that induce tradeoffs affecting performance on a given hardware, arising from intrinsic (content, e.g. scene clutter) and extrinsic (system, e.g. resource contention) characteristics. Earlier runtime execution frameworks employed rule-based decision algorithms and operated with a fixed algorithm latency budget to balance these concerns, which is sub-optimal and inflexible. We propose Chanakya, a learned approximate execution framework that naturally derives from the streaming perception paradigm, to automatically learn decisions induced by these tradeoffs instead. Chanakya is trained via novel rewards balancing accuracy and latency implicitly, without approximating either objectives. Chanakya simultaneously considers intrinsic and extrinsic context, and predicts decisions in a flexible manner. Chanakya, designed with low overhead in mind, outperforms state-of-the-art static and dynamic execution policies on public datasets on both server GPUs and edge devices.

Accurately predicting the energy consumption of an electric vehicle (EV) under real-world circumstances (such as varying road, traffic, weather conditions, etc.) is critical for a number of decisions like range estimation and route planning. A major concern for electric vehicle owners is the uncertain nature of the battery consumption. This results in the “range anxiety” and reluctance from users for mass adoption of EVs, since they are concerned about untimely drainage of battery. Even at the organizational level, a company running a fleet of electric vehicles must understand the battery consumption profiles accurately for tasks such as route and driver planning, battery sizing, maintenance planning, etc.

Billions of smoke detectors are in use worldwide to provide early warning of fires. Despite this, they frequently fail to operate in an ongoing fire, risking death and property damage. A significant fraction of faults result from drift, or reduced sensitivity, and other faults in smoke detectors' phototransistors (PTs). Existing approaches attempt to detect drift from the PT output in normal conditions (without smoke). However, we find that drifted PTs mimic the output of working PTs in normal conditions, but diverge in the presence of smoke, making this approach ineffective.

With the increasing adaption of solar energy worldwide, there is a huge interest to develop systems that help drive efficiency during manufacturing and ongoing operations. Due to various real-world conditions and processes, solar panels develop faults during their manufacturing and operations. The objective of this work is to build an End-to-End Fault Detection system to detect and localize faults in solar panels based on their Electroluminescence (EL) Imaging. Today, the majority of fault detection happens through manual inspection of EL images. To this end, we propose the design and implementation of an end-to-end system that firstly divides the solar panel into individual solar cells and then passes these cell images through a classification + detection pipeline for identifying the fault type and localizing the faults inside a cell. We propose a hybrid architecture that contains an ensemble of multiple CNN model architectures for classification and detection. The ensemble is capable of serving both - monocrystalline and polycrystalline solar panels. The proposed system significantly helps in increasing the efficiency of solar panels and reducing warranty and repair costs. We demonstrate the performance of the proposed system using an open EL image dataset with 95% of cell-level fault prediction accuracy and high recall. The proposed algorithms are applicable and can be extended for other solar applications that use RGB, EL, or thermal imaging techniques.

Information and Communication Technology (ICT) is now touching various aspects of our lives. The electricity grid with the help of ICT is transformed into Smart Grid (SG) which is highly efficient and responsive. It promotes two-way energy and information flow between energy distributors and consumers. Many consumers are becoming prosumers by also producing energy. The trend is to form small communities of consumers and prosumers leading to Micro-grids (MG) to manage energy locally. MGs are parts of SG that decentralize the energy flow by allocating the produced energy within the community. Energy allocation amongst them needs to solve issues viz., (i) how to balance supply/demand within micro-grids; (ii) how allocating energy to a user affects his/her community. To address these issues we propose six Energy Allocation Strategies (EASs) for MGs - ranging from simple to optimal. We maximize the usage of the energy generated by prosumers within MG. We form household-groups sharing similar characteristics to apply EASs by analyzing thoroughly energy and socioeconomic data of households. We propose four metrics to evaluate EASs. We test our EASs on the data from 443 households over a year. By prioritizing specific households, we increase the number of fully served households up to 81% compared to random sharing.

Fine-grained air pollution monitoring is a fundamental step towards curbing pollution levels. This is sought to be achieved by the large-scale deployment of low-cost sensors at high spatio-temporal resolution. Due to the nature of these deployments, in-the-wild and in harsh environments, sensors are prone to failures and hence ensuring data reliability is challenging. Furthermore, detecting a fault by analyzing the sensor data using existing data-centric approaches is non-trivial. This demonstration presents a sensor fault diagnosis system that employs the current signature of the sensor to address data reliability issues. The current signature captures the electrical characteristics of the hardware components enabling accurate detection and isolation of faults in low-cost pollution sensors.

Current smartphone-based navigation applications fail to provide lane-level information due to poor GPS accuracy. Detecting and tracking a vehicle’s lane position on the road assists in lane-level navigation. For instance, it would be important to know whether a vehicle is in the correct lane for safely making a turn, or whether the vehicle’s speed is compliant with a lane-specific speed limit. Recent efforts have used road network information and inertial sensors to estimate lane position. While inertial sensors can detect lane shifts over short windows, it would suffer from error accumulation over time. In this article, we present DeepLane, a system that leverages the back camera of a windshield-mounted smartphone to provide an accurate estimate of the vehicle’s current lane. We employ a deep learning--based technique to classify the vehicle’s lane position. DeepLane does not depend on any infrastructure support such as lane markings and works even when there are no lane markings, a characteristic of many roads in developing regions. We perform extensive evaluation of DeepLane on real-world datasets collected in developed and developing regions. DeepLane can detect a vehicle’s lane position with an accuracy of over 90%, and we have implemented DeepLane as an Android app.

Recent work in networked systems has shown that using aerial imagery for farm monitoring can enable precision agriculture by lowering the cost and reducing the overhead of large scale sensor deployment. However, acquiring aerial imagery requires a drone, which has high capital and operational costs, often beyond the reach of farmers in the developing world. In this paper, we present TYE (Tethered eYE), an inexpensive platform for aerial imagery. It consists of a tethered helium balloon with a custom mount that can hold a smartphone (or a camera) with a battery pack. The balloon can be carried using a tether by a person or a vehicle. We incorporate various techniques to increase the operational time of the system, and to provide actionable insights even with unstable imagery. We develop path-planning algorithms and use that to develop an interactive mobile phone application that provides the user instant feedback to guide users to efficiently traverse large areas of land. We use computer vision algorithms to stitch orthomosaics by effectively countering wind-induced motion of the camera. We have used TYE for aerial imaging of agricultural land for over a year, and envision it as a low-cost aerial imaging platform for similar applications.

Driver inattention is one of the leading causes of vehicle crashes and incidents worldwide. Driver inattention includes driver fatigue leading to drowsiness and driver distraction, say due to use of cellphone or rubbernecking, all of which leads to a lack of situational awareness. Hitherto, techniques presented to monitor driver attention evaluated factors such as fatigue and distraction independently. However, in order to develop a robust driver attention monitoring system all the factors affecting driver's attention needs to be analyzed holistically. In this paper, we present AutoRate, a system that leverages front camera of a windshield-mounted smartphone to monitor driver's attention by combining several features. We derive a driver attention rating by fusing spatio-temporal features based on the driver state and behavior such as head pose, eye gaze, eye closure, yawns, use of cellphones, etc.We perform extensive evaluation of AutoRate on real-world driving data and also data from controlled, static vehicle settings with 30 drivers in a large city. We compare AutoRate's automatically-generated rating with the scores given by 5 human annotators. Further, we compute the agreement between AutoRate's rating and human annotator rating using kappa coefficient. AutoRate's automatically-generated rating has an overall agreement of 0.87 with the ratings provided by 5 human annotators on the static dataset.

Road safety is compromised today by the inadequacies in driver license testing. Testing is typically still performed manually, and efforts aimed at automating testing are stymied by the cost of outfitting a testing track with sensors. We demonstrate a low-cost, smartphone-based system for automating key aspects of the driver license test. We have a pilot deployment of our system at an official testing track in India. We will present an analysis of license test results obtained from this pilot, comparing the smartphone-based testing results with manual evaluation.

Can a smartphone administer a driver license test? We ask this question because of the inadequacy of manual testing and the expense of outfitting an automated testing track with sensors such as cameras, leading to less-than-thorough testing and ultimately compromising road safety. We present ALT, a low-cost smartphone-based system for automating key aspects of the driver license test. A windshield-mounted smartphone serves as the sole sensing platform, with the front camera being used to monitor driver's gaze, and the rear camera, together with inertial sensors, being used to evaluate driving maneuvers such as parallel parking. The sensors are also used in tandem, for instance, to check that the driver scanned their mirror during a lane change.

T. J. Watson's apocryphal statement about there being a market for only "five computers" has, in a sense, come true with the rise of cloud computing and the dominance of a handful of "mega-computers" in terms of Internet traffic volume. However, network protocols and operation over the Internet have, for the most part, remained wedded to the old world, with individual hosts operating autonomously. We argue that this is suboptimal and that the time has come to revisit networking in the world of "five computers." We consider various networking functions, including specifically congestion control and network diagnosis, and provide an indication of the potential benefits of a new coordinated approach and sketch out an approach to realizing these benefits.

Artificial lighting is a pervasive element in our daily lives. Researchers from different communities are investigating challenges and opportunities related to artificial lighting but from different angles: energy disaggregation, to monitor the status of light bulbs in buildings; and communication, to transmit information wirelessly. We argue that there is an unexplored synergy between these two communities. When a light bulb modulates its intensity for communication, it also affects the current it draws. This current signature is unique and could be used by energy disaggregation methods to identify the lights' status. These signatures however will be exposed to interference (collisions of signatures) and distortions due to power line effects. To overcome these problems, we build upon coding schemes to assign interference-resilient signatures, and we develop custom hardware to ameliorate distortions introduced by power lines. We validate our framework in a proof-of-concept testbed, perform simulations to test scalability, and use energy traces from real homes to evaluate the impact of other electric loads.

Depletion of fossil fuel and the ever-increasing need for energy in residential and commercial buildings have triggered in-depth research on many energy saving and energy monitoring mechanisms. Currently, users are only aware of their overall energy consumption and its cost in a shared space. Due to the lack of information on individual energy consumption, users are not being able to fine-tune their energy usage. Further, even-splitting of energy cost in shared spaces does not help in creating awareness. With the advent of the Internet of Things (IoT) and wearable devices, apportioning of the total energy consumption of a household to individual occupants can be achieved to create awareness and consequently promoting sustainable energy usage. However, providing personalized energy consumption information in real-time is a challenging task due to the need for collection of fine-grained information at various levels. Particularly, identifying the user(s) utilizing an appliance in a shared space is a hard problem. The reason being, there are no comprehensive means of collecting accurate personalized energy consumption information. In this paper we present the Personalized Energy Apportioning Toolkit (PEAT) to accurately apportion total energy consumption to individual occupants in shared spaces. Apart from performing energy disaggregation, PEAT combines data from IoT devices such as smartphones and smartwatches of occupants to obtain fine-grained information, such as their location and activities. PEAT estimates energy footprint of individuals by modeling the association between the appliances and occupants in the household. We propose several accuracy metrics to study the performance of our toolkit. PEAT was exhaustively evaluated and validated in two multi-occupant households. PEAT achieves 90% energy apportioning accuracy using only the location information of the occupants. Furthermore, the energy apportioning accuracy is around 95% when both location and activity information is available.

Road safety is a major public health issue the world over. Many studies have found that the primary factors responsible for road accidents center on the driver and her/his driving. Hence, there is the need to monitor driver's state and her/his driving, with a view to providing effective feedback. Our proposed demo is of HAMS, a windshield-mounted, smartphone-based system that uses the front camera to monitor the driver and back camera to monitor her/his driving behaviour. The objective of HAMS is to provide ADAS-like functionality with low-cost devices that can be retrofitted onto the large installed base of vehicles that lack specialized and expensive sensors such as LIDAR and RADAR. Our demo would show HAMS in action on an Android smartphone to monitor the state of the driver, specifically such as drowsiness, distraction and gaze, and vehicle ranging, lane detection running on pre-recorded videos from drives.

The proliferation of Internet of Things (IoT) devices has led to the deployment of various types of sensors in the homes, offices, buildings, lawns, cities, and even in agricultural farms. Since IoT applications rely on the fidelity of data reported by the sensors, it is important to detect a faulty sensor and isolate the cause of the fault. Existing fault detection techniques demand sensor domain knowledge along with the contextual information and historical data from similar near-by sensors. However, detecting a sensor fault by analyzing just the sensor data is non-trivial since a faulty sensor reading could mimic non-faulty sensor data. This paper presents a novel primitive, which we call the Fall-curve - a sensor's voltage response when the power is turned off - that can be used to characterize sensor faults. The Fall-curve constitutes a unique signature independent of the phenomenon being monitored which can be used to identify the sensor and determine whether the sensor is correctly operating.

The proliferation of Internet of Things (IoT) devices has led to the deployment of various types of sensors in the homes, offices, buildings, lawns, cities, and even in agricultural farms. Due to the diverse nature of IoT deployments and the likelihood of sensor failures in-the-wild, a key challenge in the design of IoT systems is ensuring the integrity, accuracy, and fidelity of sensor data.

Smart grids offer better energy management at consumer premises as well as energy companies side using bi- directional communication and control. Energy companies can balance energy supply and demand to a large extent, with the advent of smart homes. They can also nudge consumers to shift their demands to off-peak hours for load balancing and monetary benefits. We propose a decentralized demand scheduling algorithm that minimizes consumer discomfort and electricity cost of a household. Our algorithm utilizes only aggregated energy consumption of a household to derive optimal appliance level demand schedules. Furthermore, a low-complexity energy disaggregation algorithm is proposed to derive fine- grained appliance information and consumer preferences. We propose three important coefficients related to energy usage of consumers. We utilize them to derive optimal day- ahead demand schedules. The decentralized algorithm is empirically evaluated using real-world energy usage data from open datasets, which include our own deployment. Our proposed scheduling algorithm saves up to 30% energy cost. This work is one of the first to derive day-ahead schedules using real-world data from multiple households.

The increase in the deployment of smart meters has enabled collection of fine-grained energy consumption data at consumer premises. Analysis of this real-time energy consumption data bestows new opportunities for better demand-response (DR) programs. This paper offers a new perspective to study energy demand and helps in designing novel mechanisms for decentralized demand-side management. Specifically, a new concept of finding the demand states using energy consumption of consumers over time and feasible transitions therein is introduced. It is shown that the orchestration of temporal transitions between the demand states can meet broad range of smart grid objectives. An online demand regulation model is developed that captures the temporal dynamics of energy demand to identify target consumers for different DR programs. This methodology is empirically evaluated and validated using data from more than 4000 households, which were part of a real-world smart grid project. This paper is the first one to comprehensively analyze the temporal dynamics of demands.

The advent of Internet of Things (IoT) has boosted the growth in number of devices around us and kindled the possibility of umpteen number of applications. One of the major challenges in the realization of IoT applications is interoperability among various IoT devices and deployments. Thus, the need for a new architecture-comprising smart control and actuation-has been identified by many researchers. In this paper, we propose a Distributed Internet-like Architecture for Things (DIAT), which will overcome most of the obstacles in the process of large-scale expansion of IoT. It specifically addresses heterogeneity of IoT devices, and enables seamless addition of new devices across applications. In addition, we propose an usage control policy model to support security and privacy in a distributed environment. We propose a layered architecture that provides various levels of abstraction to tackle the issues such as scalability, heterogeneity, security, and interoperability. The proposed architecture is coupled with cognitive capabilities that helps in intelligent decision-making and enables automated service creation. Using a comprehensive use-case, comprising elements from multiple-application domains, we illustrate the usability of the proposed architecture.

We propose energy harvesting technologies and cooperative relaying techniques to power the devices and improve reliability. We propose schemes to (a) maximize the packet reception ratio (PRR) by cooperation and (b) minimize the average packet delay (APD) by cooperation amongst nodes. Our key result and insight from the testbed implementation is about total data transmitted by each relay. A greedy policy that relays more data under a good harvesting condition turns out to be a sub optimal policy. This is because, energy replenishment is a slow process. The optimal scheme offers a low APD and also improves PRR.

The continuous growth of energy needs and the fact that unpredictable energy demand is mostly served by unsustainable (i.e. fossil-fuel) power generators have given rise to the development of Demand Response (DR) mechanisms for flattening energy demand. Building effective DR mechanisms and user awareness on power consumption can significantly benefit from fine-grained monitoring of user consumption at the appliance level. However, installing and maintaining such a monitoring infrastructure in residential settings can be quite expensive. In this paper, we study the problem of fine-grained appliance power-consumption monitoring based on one house-level meter and few plug-level meters. We explore the trade-off between monitoring accuracy and cost, and exhaustively find the minimum subset of plug-level meters that maximize accuracy. As exhaustive search is time- and resource-consuming, we define a heuristic approach that finds the optimal set of plug-level meters without utilizing any other sets of plug-level meters. Based on experiments with real data, we found that few plug-level meters-when appropriately placed — can very accurately disaggregate the total real power consumption of a residential setting and verified the effectiveness of our heuristic approach.

We implement two energy models that accurately and comprehensively estimates the system energy cost and communication energy cost for using Bluetooth and Wi-Fi interfaces. The energy models running on a system is used to smartly pick the most energy optimal network interface so that data transfer between two end points is maximized.