Find Research That Solves Problems

HarvesSink is an intelligent, under-sink greywater evaluation and diversion system designed to autonomously classify and route kitchen wastewater for safe non-potable reuse. The system integrates multi-parameter sensing across pH, Total Dissolved Solids (TDS), and turbidity with a deterministic embedded decision engine running on the STM32 microcontroller platform, ac- tuation through solenoid valves, and a fail-safe architecture that defaults to drain under any anomalous condition. The first cohort of the iPD-CP program at IIIT-Delhi established the foundational engineering of the system, progressing from ideation and design thinking through proof-of-concept (PoC) validation and onward to subsystem-level engineering design approaching TRL-5. This second semester report documents the continuation of that work, covering the complete rebuild of the hardware prototype with 12V pump-based diversion, rigorous multi-sample testing validating accuratewaterclassificationanddiversion,developmentofanEdge-AIsoftwarelayerincorporat- ing machine learning for prediction of laboratory-grade water quality metrics from inexpensive analogsensors, explorationandformalisationofanindustrialeffluenttreatmentusecase, partic- ipation in external competitions and incubation programmes, and iterative revision of the PCB schematic based on mentor feedback. The system has progressed to a stage where it performs real diversion of water samples using calibrated firmware logic, the ML layer is operational and generates predictions for Biological OxygenDemand(BOD),ChemicalOxygenDemand(COD),andAmmonia, andtheprojecthas received significant external validation through winning first prize at the Redbricks Hackathon 2025 at Ashoka University, receiving an incubation offer from IIIT-Delhi’s I-Hub Anubhuti, and entering preliminary discussions with an IIIT-Delhi incubated startup regarding the industrial ETP use case.

Democratize access to safe drinking water by evolving DetoXyFi from prototype to a commercially viable, multi-channel product.

Urban air pollution in Indian cities, particularly Delhi, presents a severe and persistent public health crisis. Particulate matter concentrations routinely exceed WHO guidelines by 10–20×, contributing to respiratory diseases, cardiovascular conditions, and reduced life expectancy. Ex- isting indoor air purifiers prioritize high single-pass filtration efficiency (H13/H14 HEPA) at the expense of air throughput, resulting in slow room-level cleaning, rapid filter clogging in high- dust environments, and high recurring maintenance costs — making them both ineffective and unaffordable for the majority of Indian households and institutions. This project, undertaken as Vayu Labs, presents the design, prototyping, and validation of an alternative indoor air purification system that inverts the conventional approach: instead of maximizing single-pass filter efficiency, the system maximizes volumetric throughput through a multi-stage filtration architecture — a washable G4 pre-filter, an activated carbon with potas- sium permanganate (ACF-K) sponge for gas-phase pollutant removal, and a medium-efficiency F6/F7 main filter. Physics-based simulation demonstrates that this 3-stage system achieves a Clean Air Delivery Rate (CADR) of 634 m3/hr with a t 90 of 3.95 minutes in an 18.1 m3 room — 42% higher CADR and 29% faster room cleaning than a single-stage H14 HEPA system — at 79% lower annual filter cost (≈INR 11,553 vs. INR 55,233). The project combines hardware engineering (airflow system design, filter selection, sensor in- tegration, enclosure CAD) with entrepreneurial validation (customer discovery, competitive analysis, institutional go-to-market strategy, and pitch competition participation). The system targets both residential consumers (INR 8,000–10,000 retail) and institutional buyers (schools, government offices) where the cost-per-CADR advantage is most compelling. A seed grant of INR 1,00,000 was secured through the E-Pitch Caf´e 10th Edition (Budding Star category) to fund prototype fabrication and pilot deployments. Field of Study: Environmental Engineering and Sustainable Development (aligned with UN SDG 9: Industry & Innovation, SDG 11: Sustainable Cities, and SDG 13: Climate Action)

HarvesSink is an intelligent, under-sink greywater evaluation and diversion system designed to autonomously classify and route kitchen wastewater for safe non-potable reuse. The system integrates multi-parameter sensing across pH, Total Dissolved Solids (TDS), and turbidity with a deterministic embedded decision engine running on the STM32 microcontroller platform, ac- tuation through solenoid valves, and a fail-safe architecture that defaults to drain under any anomalous condition. The first cohort of the iPD-CP program at IIIT-Delhi established the foundational engineering of the system, progressing from ideation and design thinking through proof-of-concept (PoC) validation and onward to subsystem-level engineering design approaching TRL-5. This second semester report documents the continuation of that work, covering the complete rebuild of the hardware prototype with 12V pump-based diversion, rigorous multi-sample testing validating accuratewaterclassificationanddiversion,developmentofanEdge-AIsoftwarelayerincorporat- ing machine learning for prediction of laboratory-grade water quality metrics from inexpensive analogsensors, explorationandformalisationofanindustrialeffluenttreatmentusecase, partic- ipation in external competitions and incubation programmes, and iterative revision of the PCB schematic based on mentor feedback. The system has progressed to a stage where it performs real diversion of water samples using calibrated firmware logic, the ML layer is operational and generates predictions for Biological OxygenDemand(BOD),ChemicalOxygenDemand(COD),andAmmonia, andtheprojecthas received significant external validation through winning first prize at the Redbricks Hackathon 2025 at Ashoka University, receiving an incubation offer from IIIT-Delhi’s I-Hub Anubhuti, and entering preliminary discussions with an IIIT-Delhi incubated startup regarding the industrial ETP use case.

Precisenutritionaltrackingisfoundationaltohealth,yetmanualandimage-onlymethodsare tediousandinaccurate. ThispaperintroducesSmartScale,aconsumerIoTdevicepairedwith FoodSense, a native mobile wellness platform that overcomes these challenges by integrating high-precisionweightsensingwithon-devicemultimodalAI.SmartScalecombinesacalibrated HX711 load cell on an ESP32-WROOM with FoodSense’s dual-backend inference pipeline: (i)Google’sGemma4E2B(Q4 K M,3.2GB)executedentirelyon-deviceviallama.cppwith MetalGPU acceleration, (ii)aYOLOv8TFLite detectorfine-tunedon94+ Indianfoodclasses, and(iii)anoptionalGeminiCloudproviderforuserswhosupplytheirownkey. Thesystemis shipped as a software under the Apache 2.0 license and ensures that food images never leave the user’s device by default. This work details the hardware, embedded, and on-device AI architecture anddemonstrates its potential toprovide reliable, private, subscription-freedietary monitoring.

Since 2016, when the Facebook/Cambridge Analytica scandal began to emerge, public concern has grown around the threat of “online manipulation”. While these worries are familiar to privacy researchers, this paper aims to make them more salient to policymakers—first, by defining “online manipulation”, thus enabling identification of manipulative practices; and second, by drawing attention to the specific harms online manipulation threatens. We argue that online manipulation is the use of information technology to covertly influence another person’s decision-making, by targeting and exploiting their decision-making vulnerabilities. Engaging in such practices can harm individuals by diminishing their economic interests, but its deeper, more insidious harm is its challenge to individual autonomy. We explore this autonomy harm, emphasising its implications for both individuals and society, and we briefly outline some strategies for combating online manipulation and strengthening autonomy in an increasingly digital world.

Urban air pollution in Indian cities, particularly Delhi, presents a severe and persistent public health crisis. Particulate matter concentrations routinely exceed WHO guidelines by 10–20×, contributing to respiratory diseases, cardiovascular conditions, and reduced life expectancy. Ex- isting indoor air purifiers prioritize high single-pass filtration efficiency (H13/H14 HEPA) at the expense of air throughput, resulting in slow room-level cleaning, rapid filter clogging in high- dust environments, and high recurring maintenance costs — making them both ineffective and unaffordable for the majority of Indian households and institutions. This project, undertaken as Vayu Labs, presents the design, prototyping, and validation of an alternative indoor air purification system that inverts the conventional approach: instead of maximizing single-pass filter efficiency, the system maximizes volumetric throughput through a multi-stage filtration architecture — a washable G4 pre-filter, an activated carbon with potas- sium permanganate (ACF-K) sponge for gas-phase pollutant removal, and a medium-efficiency F6/F7 main filter. Physics-based simulation demonstrates that this 3-stage system achieves a Clean Air Delivery Rate (CADR) of 634 m3/hr with a t 90 of 3.95 minutes in an 18.1 m3 room — 42% higher CADR and 29% faster room cleaning than a single-stage H14 HEPA system — at 79% lower annual filter cost (≈INR 11,553 vs. INR 55,233). The project combines hardware engineering (airflow system design, filter selection, sensor inte- gration, enclosure CAD) with entrepreneurial validation (customer discovery, competitive anal- ysis, institutional go-to-market strategy, and pitch competition participation). The system targets both residential consumers (INR 8,000–10,000 retail) and institutional buyers (schools, government offices) where the cost-per-CADR advantage is most compelling. A seed grant of INR 1,00,000 was secured through the E-Pitch Café 10th Edition (Budding Star category) to fund prototype fabrication and pilot deployments.

Fuelled by increasing computer power and algorithmic advances, ma- chine learning techniques have become powerful tools for finding patterns in data. Since quantum systems produce counter-intuitive patterns be- lieved not to be efficiently produced by classical systems, it is reasonable to postulate that quantum computers may outperform classical comput- ers on machine learning tasks. The field of quantum machine learning explores how to devise and implement concrete quantum software that offers such advantages. Recent work has made clear that the hardware and software challenges are still considerable but has also opened paths towards solutions.

3 Acknowledgment 4

3 Acknowledgments 4

Collisions at non-line-of-sight (NLOS) blind curves are a persistent and fatal problem, particularly in mountainous regions where conventional safety measures are either unre- liable (convex mirrors) or prohibitively expensive (radar systems). This project presents theBlind Turn Accident Alert System (BTAAS),alow-cost,solar-poweredVehicle- to-Infrastructure (V2I) product ecosystem designed to bridge this safety gap. The system comprises two key hardware components: a dashboard-mounted Vehicle Node (VN) and an intelligent solar-powered Pole Node (PN). Using Sub-GHz LoRa technology, the system establishes a digital line-of-sight around curves. To solve the critical challenge of false alarms, we developed ”Vector-Track,” a novel IMU-based direc- tional sensing algorithm that utilizes quaternion mathematics to distinguish between safe (same-direction) and hazardous (converging) traffic with 100% accuracy in field trials. Experimental validation confirmed a reliable link at 150m (RSSI -91.3 dBm). The power architecture ensures 3.2 days of autonomy using a 10W solar panel and LiFePO4 storage. With a unit cost of approx. Rs. 1,000 per vehicle node, BTAAS offers a scalable alternative to radar for the accident-prone curves in the mountainous regions of India.

. We prove a result analogous to Reeb’s theorem in the context of Morse-Bott functions: if a closed, smooth manifold M admits a Morse-Bott function having two critical submanifolds Sk and Sl (k̸=l), then M hasdimensionk+l+1andishomeomorphictothestandardsphereSk+l+1 butnotnecessarilydiffeomorphicto it. Wealsoprovesimilarresultsforprojectivespacesovertherealnumbers,complexnumbersandquaternions.