From what to build to how to build it — our platform helps you understand, plan, and solve SIH 2026 problem statements with complete insights into requirements, solution approaches, technology stacks, architecture, AI/ML, and implementation roadmaps.
Search for your problem statement, and start building your solution.
| Description: | • Background The government actively supports the socio-economic upliftment of marginalized communities,particularly micro-entrepreneurs, artisans, and weavers. Financial assistance is provided to establish small-scale manufacturing and handicraft units. To help these beneficiaries sell their goods, market exposure is facilitated through periodic physical exhibitions, cluster development programs, and trade fairs (such as Shilp Samagam, Surajkund Mela, and Dilli Haat).While physical exhibitions provide a temporary boost in sales, these micro-entrepreneurs lack continuous, year-round access to broader digital markets. Transitioning to the digital economy is hindered by low digital literacy, language barriers, and a lack of technical skills required to professionally photograph, price, and catalog products for modern e-commerce. • Challenge There is a critical need to bridge the gap between traditional craftsmanship and modern digital commerce. Beneficiaries struggle to present their products competitively online. They often fail to capture high-quality images, write compelling product descriptions, or understand dynamic market pricing.The challenge is to build an intuitive, AI-driven mobile application that acts as a 'virtual business manager' for these artisans. The app must empower them to seamlessly digitize their inventory, optimize their listings using AI, and connect directly with larger B2B buyers or government e-marketplaces without requiring advanced technical knowledge. • Expected Solution Participants are expected to develop an AI-powered, cross-platform mobile application supported by a robust, scalable backend architecture. To ensure high adoption among low-literacy users, the application must feature a highly responsive, minimalist UI/UX design (incorporating modern, clean visual hierarchies and accessible layouts). Key features should include: 1. AI Image Enhancer & Studio: A built-in camera module that utilizes AI to automatically remove cluttered backgrounds, correct lighting, and format product photos (e.g., textiles,handicrafts) to professional e-commerce standards. 2. Multilingual Auto-Cataloger: An NLP-based engine that allows artisans to describe their product via voice notes in regional languages. The AI should translate and generate SEO-friendly, professional product descriptions in English and Hindi. 3. Dynamic Pricing Assistant: A machine learning algorithm that analyzes the uploaded product image and description to suggest an optimal, competitive selling price based on current market trends and raw material costs. • Impact Goals • Provide marginalized micro-entrepreneurs with a continuous, year-round digital sales channel,reducing their dependency on periodic physical fairs. • Drastically lower the barrier to entry for digital commerce through intuitive AI automation. • Improve digital literacy and financial independence, ultimately increasing the average annual income of the target demographic. |
| Organization: | Ministry of Social Justice and Empowerment (MoSJE) |
| Department: | Department of Social Justice and Empowerment |
| Category: | Software |
| Theme: | Heritage & Culturetd> |
| Youtube Link: | Not provided |
| Dataset Link: | Not provided |
| Contact info: | Not provided |
Marginalized artisans, weavers, and micro-entrepreneurs supported by the Ministry of Social Justice and Empowerment (MoSJE) face severe economic stagnation due to reliance on seasonal physical exhibitions like Shilp Samagam and Dilli Haat [PS]. The transition to year-round digital commerce is blocked by low digital literacy, language barriers, and an inability to create professional e-commerce product listings [PS]. This dossier presents CraftVeda AI (working title), a cross-platform mobile application and scalable microservices backend that serves as an intelligent "virtual business manager" for rural artisans [PS]. The solution integrates on-device AI image enhancement (background removal, light correction) [PS, RESEARCHED], multilingual voice-to-cataloging via Government of India's Bhashini engine [PS, RESEARCHED], ML-driven price recommendation considering market trends and material costs [PS, INFERENCE], and direct multi-channel dispatch to the Open Network for Digital Commerce (ONDC) and Government e-Marketplace (GeM) [PS, RESEARCHED]. Designed for low-bandwidth, low-literacy environments [PS, INFERENCE], this blueprint provides a student hackathon team with an actionable roadmap to build an MVP for SIH 2026.
| Parameter | Value |
|---|---|
| Problem Statement ID | SIH26090 [PS] |
| Title | AI-Driven Market Linkage and Smart Cataloging Mobile Application for Marginalized Artisans [PS] |
| Organization | Ministry of Social Justice and Empowerment (MoSJE) [PS] |
| Department | Department of Social Justice and Empowerment [PS] |
| Category | Software [PS] |
| Theme | Heritage & Culture [PS] |
| Submission Deadline | 30 September 2026 [PS] |
| Idea Submissions Counter | 1/500 [PS] |
| Fundamental System Objective | To digitize artisan inventory using voice-first AI and computer vision, and automatically publish optimized listings to national digital commerce networks (ONDC/GeM) to enable continuous, year-round income [PS, INFERENCE]. |
The problem statement has been extracted into atomic, non-collapsible requirements, preserving all original intent:
REQ-1: Cross-Platform Mobile Application
REQ-2: Minimalist UI/UX for Low Digital Literacy
REQ-3: Key Feature 1 - AI Image Enhancer & Studio
REQ-4: Key Feature 2 - Multilingual Auto-Cataloger
REQ-5: Key Feature 3 - Dynamic Pricing Assistant
REQ-6: Direct Market Linkage (B2B & Government e-Marketplaces)
| Stakeholder | Goal | Current Pain | System Need |
|---|---|---|---|
| Marginalized Artisan / Weaver | Earn steady year-round income from handmade products [PS]. | Income is restricted to periodic physical fairs (Shilp Samagam, Surajkund) [PS]; low digital literacy, language barriers [PS]. | Ultra-simple, voice-first mobile app with automated camera studio and instant listing creation [PS, INFERENCE]. |
| MoSJE & Apex Corporations (NSFDC, NBCFDC, NSKFDC, TULIP) | Fulfill mandates for socio-economic upliftment of SC/OBC/Safai Karamchari artisans [RESEARCHED]. | Lack of real-time operational visibility into artisan sales and inventory post-exhibitions [INFERENCE]. | Central monitoring dashboard, aggregate analytics, and verified beneficiary onboarding verification [INFERENCE]. |
| B2B / Wholesale Buyers & Retail Consumers | Source authentic handmade Indian handicrafts at transparent prices [INFERENCE]. | Unstandardized product photos, missing technical specifications, poor price discoverability [INFERENCE]. | High-resolution studio product photos, standardized multilingual descriptions, and fair market pricing [PS, INFERENCE]. |
| Cluster Facilitators / SHG Leads | Assist group members in bulk cataloging and order handling [INFERENCE]. | Time-consuming manual data entry and photography for dozens of artisans [INFERENCE]. | Batch cataloging mode, delegated account management, and consolidated fulfillment view [INFERENCE]. |
| Feature / Subsystem | Justifying Requirement | Technical Scope & Implementation |
|---|---|---|
| Guided Camera Module with On-Device Auto-Crop | REQ-3 (AI Image Enhancer) [PS] | Custom Flutter camera overlay with edge detection and automatic background isolation using ONNX Runtime [RESEARCHED]. |
| AI Studio Background Transformer & Relighting | REQ-3 (AI Image Enhancer) [PS] | BRIA RMBG-1.4 background removal + CLAHE lighting equalization microservice [RESEARCHED]. |
| Voice-to-Text Regional Audio Recorder | REQ-4 (Multilingual Auto-Cataloger) [PS] | Audio capture in 12+ Indian regional languages, streamed to Bhashini STT API [RESEARCHED]. |
| NLP Structured Description & SEO Generator | REQ-4 (Multilingual Auto-Cataloger) [PS] | Prompt-engineered LLM microservice converting raw transcripts into bilingual (Hindi/English) bullet points and search tags [INFERENCE]. |
| Multimodal Price Estimation Engine | REQ-5 (Dynamic Pricing Assistant) [PS] | XGBoost + CLIP embedding regression model taking raw material input and visual feature classification [RESEARCHED, INFERENCE]. |
| ONDC & GeM Protocol Gateway Adapter | REQ-6 (Direct Market Linkage) [PS] | Beckn protocol implementation (search, select, init, confirm) for publishing inventory to ONDC buyer apps [RESEARCHED]. |
| Voice-Guided Minimalist Flutter Frontend | REQ-2 (Minimalist UI/UX) [PS] | Single-action UI, audio tooltips in regional dialects, large high-contrast visual cues [INFERENCE]. |
| Data Type | Purpose | Ideal Real-World Source | Hackathon Substitute | Access Method | Availability | Privacy Sensitivity | Confidence Tag |
|---|---|---|---|---|---|---|---|
| Handicraft Catalog Images | Train & test AI studio image enhancer | MoSJE TULIP brand portal, Tribes India catalog [RESEARCHED] | Public Kaggle Handicraft datasets + custom collected 100 craft samples | Direct download / Scraping | Publicly available | Low | [RESEARCHED] |
| Regional Audio Samples | Test voice-to-text in regional languages | AI4Bharat / Bhashini Universal Language Corpus [RESEARCHED] | Custom recorded voice clips in 4 languages (Hindi, Marathi, Tamil, Bengali) | Bhashini Open API / Local WAV upload | Public API | Low | [RESEARCHED] |
| Handicraft Market Pricing Data | Train ML Dynamic Pricing model | GeM public procurement transaction logs, ONDC retail price benchmarks [RESEARCHED] | Curated synthetic dataset of 1,000 craft listings (Craft Type, Material, Labor Hours, Market Price) | CSV / PostgreSQL import | Synthetic / Researched | Low | [ASSUMPTION] |
| MoSJE Beneficiary Records | Validate artisan identity & corporation affiliation | NBCFDC / NSFDC beneficiary databases [RESEARCHED] | Mock JSON database matching Beneficiary ID format | REST API / Mock Service | Simulated | High (PII) | [ASSUMPTION] |
| ONDC Product Category Mapping | Map craft items to standardized ONDC taxonomy | ONDC Retail v1.2.0 Category Taxonomy [RESEARCHED] | Official ONDC JSON Schema taxonomy file | GitHub repository clone | Open Source | None | [RESEARCHED] |
+-------------------------------------------------------------------------------------------------------------------------+
| CANDIDATE SOLUTION APPROACHES |
+------------------------------------+------------------------------------+-----------------------------------------------+
| Approach 1: Cloud-Centric Generative| Approach 2: Lightweight Hybrid AI | Approach 3: Rule-Based & Offline Studio App |
| AI Microservice Architecture | Microservices (RECOMMENDED) | (Edge-Only) |
+------------------------------------+------------------------------------+-----------------------------------------------+
| - Full Cloud Processing | - On-Device Capture & Light Seg. | - Local Open-CV thresholding |
| - Heavy LLMs (GPT-4o / Claude) | - Server-Side RMBG-1.4 / Bhashini | - Static template descriptions |
| - Complex Multimodal APIs | - XGBoost Dynamic Pricing | - Fixed Cost-Plus Math Pricing |
| - High latency & recurring cost | - Optimized for low-spec phones | - Zero AI/ML maintenance cost |
+------------------------------------+------------------------------------+-----------------------------------------------+
| Evaluation Criteria | Approach 1: Cloud-Centric Heavy AI | Approach 2: Lightweight Hybrid AI (RECOMMENDED) | Approach 3: Edge-Only Rule-Based |
|---|---|---|---|
| Problem Fit | High capability, but over-engineered for low network [INFERENCE]. | High fit; balances offline mobile capture with scalable cloud microservices [INFERENCE]. | Low fit; fails requirement for professional AI image studio & dynamic pricing [INFERENCE]. |
| Accuracy Potential | High image generation quality, but risk of LLM hallucinations [INFERENCE]. | High precision using specialized RMBG-1.4, Bhashini STT, and trained tabular XGBoost [RESEARCHED]. | Low; rule-based background removal fails on complex handicraft textures [INFERENCE]. |
| Data Requirements | Requires continuous high-bandwidth internet upload [INFERENCE]. | Modest; compressed audio and image payloads sent via REST [INFERENCE]. | Zero cloud data required; local rules only [INFERENCE]. |
| Complexity | High operational cost & API latency [INFERENCE]. | Moderate; clean modular microservices [INFERENCE]. | Low system complexity [INFERENCE]. |
| Explainability | Low (Black box LLM outputs) [INFERENCE]. | High; XGBoost feature importance score & transparent pricing breakdown [INFERENCE]. | Deterministic (100% explainable) [INFERENCE]. |
| Cost | Unviable long-term (High API tokens) [INFERENCE]. | Low (Uses free Bhashini APIs & self-hosted open-source models) [RESEARCHED]. | Very Low [INFERENCE]. |
| Scalability | Cloud bottle-necked [INFERENCE]. | Highly scalable microservice container architecture [INFERENCE]. | Limited by mobile hardware processing speed [INFERENCE]. |
| Hackathon Feasibility | Fast to prototype, fragile in live presentation [INFERENCE]. | Excellent buildability within student hackathon scope [INFERENCE]. | Easy to build, fails judging evaluation criteria [INFERENCE]. |
| Production Feasibility | High API cost burden for MoSJE [INFERENCE]. | Production ready; aligns with Digital India stack (Bhashini, ONDC) [RESEARCHED]. | Unsuitable for modern e-commerce standards [INFERENCE]. |
Selected Approach: Approach 2: Lightweight Hybrid AI Microservices Architecture [INFERENCE, ASSUMPTION].
+---------------------------------------------------------------------------------------------------+
| CLIENT LAYER (CROSS-PLATFORM MOBILE APP) |
| [Flutter Android/iOS App] -- [Minimalist Voice UI] -- [Custom Camera Module] -- [SQLite Sync Cache] |
+---------------------------------------------------------------------------------------------------+
|
HTTPS / REST / WebSockets / WSS
v
+---------------------------------------------------------------------------------------------------+
| API GATEWAY & SECURITY LAYER |
| [Nginx Reverse Proxy] -- [JWT Auth / Firebase Auth] -- [Rate Limiter & Request Sanitizer] |
+---------------------------------------------------------------------------------------------------+
|
+------------------------------------------+------------------------------------------+
| | |
v v v
+--------------------------+ +--------------------------+ +----------------------------------+
| AI IMAGE STUDIO SERVICE | | MULTILINGUAL NLP SERVICE | | DYNAMIC PRICING SERVICE |
| - BRIA RMBG-1.4 (ONNX) | | - Bhashini STT Engine | | - XGBoost Regressor |
| - OpenCV Light Correction| | - Bhashini NMT Engine | | - CLIP Image Feature Extractor |
| - Auto Aspect Formatter | | - Llama-3 / Mistral LLM | | - Market Price Benchmark DB |
+--------------------------+ +--------------------------+ +----------------------------------+
| | |
+------------------------------------------+------------------------------------------+
|
v
+---------------------------------------------------------------------------------------------------+
| CORE BACKEND & PERSISTENCE LAYER |
| [FastAPI Framework] <---> [PostgreSQL + PostGIS] <---> [Redis Cache / Queue] <---> [AWS S3 / MinIO] |
+---------------------------------------------------------------------------------------------------+
|
v
+---------------------------------------------------------------------------------------------------+
| EXTERNAL INTEGRATION LAYER |
| [ONDC Beckn Protocol Gateway (ISN)] <---> [GeM Catalog Adapter] <---> [MoSJE Beneficiary API] |
+---------------------------------------------------------------------------------------------------+
Step 1: Capture & Offline Buffer The artisan opens CraftVeda AI, selects their native language via voice prompt, and taps a single giant button to snap a product photo [INFERENCE]. The image is downscaled to 1080p and stored in local SQLite along with a raw audio voice note detailing craft type, material, and creation time [INFERENCE].
Step 2: Secure Transmission When network connectivity is detected, the background sync service pushes the binary payload over HTTPS (TLS 1.3) with JWT bearer token authentication to the API Gateway [INFERENCE].
Step 3: Parallel AI Processing
Step 4: Dynamic Price Computation The Dynamic Pricing microservice combines extracted material cost and labor hours with visual complexity vectors from CLIP [RESEARCHED]. The XGBoost engine queries historical marketplace price distributions for similar craft categories and outputs a recommended price (e.g., ₹1,450) with an adjustable range slider (₹1,200 – ₹1,700) [INFERENCE].
Step 5: Voice Feedback & Validation The system speaks back the generated title, description, and suggested price in the artisan's native language using Bhashini TTS [RESEARCHED]. The artisan confirms approval with a single voice command ("Sahi Hai" / "Confirm") [INFERENCE].
Step 6: Network Dispatch The core service updates PostgreSQL, uploads studio images to MinIO/S3, and formats the listing into Beckn protocol JSON payloads, broadcasting the inventory to ONDC buyer apps and pushing candidate catalogs to the GeM portal [RESEARCHED].
| Layer | Selected Technology | Why It Fits | Alternative Considered | Why Rejected |
|---|---|---|---|---|
| Frontend Mobile | Flutter (Dart) | Single codebase for Android/iOS; superior native performance for camera/audio handling; lightweight UI renders cleanly on budget phones [RESEARCHED]. | React Native | Slower native bridge performance when handling camera streams and raw ONNX runtime execution [INFERENCE]. |
| API Gateway | Nginx + FastAPI | High-throughput asynchronous Python framework; native integration with ML pipelines; automatic OpenAPI docs [KNOWN]. | Node.js (Express) | Python FastAPI handles native PyTorch/ONNX model inference bindings far more efficiently [INFERENCE]. |
| Database | PostgreSQL + PostGIS | Robust relational model for multi-tenant seller inventory; GIS spatial queries for mapping artisan clusters across Indian districts [KNOWN]. | MongoDB | Relational integrity required for multi-channel order settlement and financial reporting [INFERENCE]. |
| Caching & Queues | Redis + Celery | Asynchronous task processing for heavy AI model pipelines; prevents HTTP request timeouts during image processing [KNOWN]. | RabbitMQ | Redis provides combined caching and pub/sub queueing with smaller memory footprint [INFERENCE]. |
| AI - Background Removal | BRIA RMBG-1.4 (ONNX) | State-of-the-art open background removal model; runs via ONNX Runtime CPU/GPU with <2 second latency [RESEARCHED]. | remove.bg API | Commercial API costs ($0.20/image) are unviable for hackathon scale and production deployment [RESEARCHED]. |
| AI - Multilingual Speech | Digital India Bhashini API | Official GoI public infrastructure supporting STT, NMT, and TTS across 22 scheduled Indian languages [RESEARCHED]. | OpenAI Whisper API | Whisper lacks fine-tuned accuracy for hyper-local Indian regional accents and costs per audio minute [RESEARCHED]. |
| AI - Dynamic Pricing | XGBoost + OpenCV/CLIP | Interpretable gradient boosting on structured metadata (material, hours) fused with visual embeddings [RESEARCHED]. | Deep Neural Nets (MLP) | Deep networks require massive training datasets and act as unexplainable black boxes for pricing [INFERENCE]. |
| Storage | AWS S3 / MinIO | Industry standard object storage for studio assets; MinIO allows local self-hosting during hackathon build [KNOWN]. | Firebase Storage | Vendor lock-in; higher cost at scale compared to MinIO/S3 [INFERENCE]. |
| Market Integration | ONDC Seller App SDK (Beckn) | Official open-source protocol SDK for publishing inventory directly onto the ONDC network [RESEARCHED]. | Proprietary APIs | Closed e-commerce integrations defeat the problem objective of open market access [PS, INFERENCE]. |
-- Artisan / Beneficiary Table
CREATE TABLE artisans (
artisan_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
mosje_beneficiary_id VARCHAR(50) UNIQUE NOT NULL,
full_name VARCHAR(100) NOT NULL,
phone_number VARCHAR(15) UNIQUE NOT NULL,
state VARCHAR(50) NOT NULL,
district VARCHAR(50) NOT NULL,
preferred_language VARCHAR(10) DEFAULT 'hi',
corporation_type VARCHAR(20) CHECK (corporation_type IN ('NSFDC', 'NBCFDC', 'NSKFDC', 'OTHER')),
created_at TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
);
-- Product Inventory Table
CREATE TABLE products (
product_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
artisan_id UUID REFERENCES artisans(artisan_id) ON DELETE CASCADE,
raw_audio_url TEXT,
original_image_url TEXT NOT NULL,
studio_image_url TEXT,
title_hi VARCHAR(200),
title_en VARCHAR(200),
description_hi TEXT,
description_en TEXT,
craft_category VARCHAR(100) NOT NULL,
material_cost NUMERIC(10, 2),
labor_hours NUMERIC(5, 1),
recommended_price NUMERIC(10, 2),
final_price NUMERIC(10, 2) NOT NULL,
stock_quantity INT DEFAULT 1,
ondc_item_id VARCHAR(100),
sync_status VARCHAR(20) DEFAULT 'DRAFT' CHECK (sync_status IN ('DRAFT', 'PENDING_SYNC', 'PUBLISHED', 'FAILED')),
created_at TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
);
-- Local SQLite Schema (Mobile Device Storage)
CREATE TABLE local_drafts (
draft_id TEXT PRIMARY KEY,
local_image_path TEXT NOT NULL,
local_audio_path TEXT,
craft_category TEXT,
material_cost REAL,
sync_state INTEGER DEFAULT 0 -- 0: Unsynced, 1: Syncing, 2: Synced
);
WorkManager to trigger an exponential backoff retry sync whenever an active internet connection (Wi-Fi or cellular) is detected [INFERENCE, ASSUMPTION].+---------------------------------------------------------------------------------------------------+
| HACKATHON MVP SCOPE |
+-------------------------------------------------+-------------------------------------------------+
| IN SCOPE | DESCRIBED / OUT OF SCOPE |
| (To Build & Demo) | (Future Production) |
+-------------------------------------------------+-------------------------------------------------+
| [x] Flutter Android Mobile Application | [ ] Production iOS App Store Release |
| [x] Custom Camera & AI Background Remover | [ ] Direct Hardware Edge Neural Accelerator |
| [x] Multilingual Voice Recording (4 Dialects) | [ ] Full 22 Language Real-Time Voice Models |
| [x] Bhashini Integration (STT/NMT/TTS) | [ ] Real-Time Payment Gateway Escrow Settlement |
| [x] Dynamic Price Recommendation Model | [ ] Live GeM Portal Production Vendor Approval |
| [x] ONDC Mock Sandbox Seller Node Integration | [ ] Automated Multi-State Logistics Dispatch |
| [x] MoSJE Admin Monitoring Web Dashboard | [ ] Production Security Audit & Certifications |
+-------------------------------------------------+-------------------------------------------------+
To ensure a flawless live presentation during SIH judging without relying on flaky live network conditions or external API limits:
Simulated Bhashini Voice Pipeline:
Simulated ONDC Network Buyer View:
Simulated Beneficiary Verification:
NBCFDC-2026-9941) pre-populated with demographic and craft cluster data [ASSUMPTION].Aligned against the SIH deadline of 30 September 2026 [PS]:
+---------------------------------------------------------------------------------------------------+
| IMPLEMENTATION ROADMAP (SIH 2026) |
+---------------------------------------------------------------------------------------------------+
| PHASE 1: Foundations & AI Models [Target: July 15, 2026] |
| - Set up Flutter project & clean architecture. |
| - Deploy RMBG-1.4 & OpenCV pipeline on FastAPI backend. |
| - Curate 1,000 craft pricing dataset & train XGBoost model. |
+---------------------------------------------------------------------------------------------------+
|
v
+---------------------------------------------------------------------------------------------------+
| PHASE 2: Multilingual Voice & Core API [Target: August 15, 2026] |
| - Integrate Bhashini STT/NMT/TTS APIs. |
| - Connect LLM structured output parser for catalog generation. |
| - Implement PostgreSQL schema & Redis task queues. |
+---------------------------------------------------------------------------------------------------+
|
v
+---------------------------------------------------------------------------------------------------+
| PHASE 3: UI Polish, ONDC Sandbox & Testing [Target: September 10, 2026] |
| - Implement minimalist Flutter UI with native voice feedback. |
| - Integrate ONDC Beckn protocol seller gateway adapter. |
| - Execute field testing with simulated low-literacy users. |
+---------------------------------------------------------------------------------------------------+
|
v
+---------------------------------------------------------------------------------------------------+
| PHASE 4: Final Demo Freeze & Submission [Target: September 25, 2026] |
| - Freeze code, build APK/IPA binaries. |
| - Set up fallback demo simulation scripts & judging presentation deck. |
| - Final Submission ahead of September 30, 2026 deadline [PS]. |
+---------------------------------------------------------------------------------------------------+
| Test Category | Target Metric / Requirement | Test Methodology |
|---|---|---|
| Functional Testing | 100% successful execution of Capture -> Studio -> Voice -> Price -> Publish flow [INFERENCE]. | Automated end-to-end integration tests using Flutter Driver & PyTest [INFERENCE]. |
| AI Image Quality | Background removal accuracy mIoU > 0.90 across 50 craft types [RESEARCHED]. | Benchmark against dataset of raw craft photos taken in cluttered rural environments [INFERENCE]. |
| Voice Recognition | Speech-to-Text Word Error Rate (WER) < 15% in target languages [RESEARCHED]. | Test against regional dialect audio clips recorded near noisy handloom looms [INFERENCE]. |
| Dynamic Pricing | Mean Absolute Percentage Error (MAPE) < 12% against market prices [RESEARCHED]. | K-fold cross-validation on craft market pricing dataset [INFERENCE]. |
| Usability Testing | Task completion time < 2 minutes for a low-literacy user [INFERENCE]. | Usability testing with 5 non-tech savvy users using voice guidance only [INFERENCE]. |
| Offline Resilience | Zero data loss during abrupt network disconnect mid-upload [INFERENCE]. | Network throttling & forced disconnection testing using Android Emulator [INFERENCE]. |
| Risk Description | Likelihood | Impact | Severity | Mitigation Strategy |
|---|---|---|---|---|
| Bhashini API Downtime / Latency | Medium | High | High | Implement local Whisper-Tiny / Edge translation fallback model in FastAPI microservice [RESEARCHED, ASSUMPTION]. |
| Low Network Speeds in Rural Clusters | High | High | High | Compress images to WebP (<150KB) on device before upload; full offline queueing in SQLite [INFERENCE]. |
| Artisan Hesitation to Accept AI Price | Medium | Medium | Medium | Display price transparently with material/labor break-up; grant 100% manual price override control to artisan [INFERENCE]. |
| Complex Craft Textures Destroyed by BG Removal | Medium | High | High | Add manual "Background Color Selection" toggle and adjustable edge-feathering threshold in app [RESEARCHED, INFERENCE]. |
| ONDC Sandbox Integration Protocol Changes | Low | High | Medium | Containerize Beckn protocol adapter using official ONDC Seller App SDK [RESEARCHED]. |
+---------------------------------------------------------------------------------------------------+
| DIFFERENTIATION MATRIX |
+-----------------------------------+-----------------------------------+---------------------------+
| Generic E-Commerce Submission | Traditional Government Portals | CraftVeda AI (OUR SYSTEM) |
+-----------------------------------+-----------------------------------+---------------------------+
| - Text-heavy forms | - Desktop-first web portals | - Voice-first UI |
| - English/Hindi script only | - Manual photo uploads required | - 12+ Regional Dialects |
| - Standard camera upload | - Fixed manual pricing | - AI Image Studio |
| - Closed marketplace model | - Siloed database | - ML Dynamic Pricing |
| | | - Direct ONDC/GeM Linkage |
+-----------------------------------+-----------------------------------+---------------------------+
Proceed with building CraftVeda AI using Approach 2 (Lightweight Hybrid AI Microservices) [INFERENCE, ASSUMPTION]. This solution directly addresses every requirement in problem statement SIH26090 while maintaining high feasibility for a 4-6 student team within the deadline of 30 September 2026 [PS].
Access to active Bhashini REST/gRPC API credentials for regional language Speech-To-Text and Translation [RESEARCHED]. Fallback: Self-hosted Whisper-Tiny + IndicTrans2 microservice.
Intermittent rural internet connectivity during artisan photo/audio upload [PS, INFERENCE]. Mitigation: Full offline local SQLite queueing and WebP/AAC binary compression.
Achieving high-accuracy background removal on intricate handicraft textures (e.g., chikankari lace, thin wooden carvings, handloom threads) without destroying product edges [INFERENCE].
92 / 100 [INFERENCE] (High score justified by strong availability of open-source models like RMBG-1.4, official Bhashini public APIs, and standardized Flutter/FastAPI tech stack) [RESEARCHED].
High (90%) [INFERENCE] (Plan is fully grounded in the supplied problem JSON, supported by verified public digital infrastructure like Bhashini and ONDC) [PS, RESEARCHED].
Title: Nomination of the Artisans for Exhibitions/Melas - NBCFDC [1.1]
https://nbcfdc.gov.in/nbcfdc/web/en/Nomination-of-the-Artisans-for-ExhibitionsTitle: Union Minister Dr. Virendra Kumar inaugurates 'TULIP' brand - A Digital Platform for Craft and Community Empowerment [1.2, 1.7, 1.9]
https://www.pib.gov.in/PressReleasePage.aspx?PRID=2070951Title: Bhashini AI - Making Languages More Accessible with Digital Technology [2.2, 2.3]
https://www.unicef.org/digitalimpact/bhashini-ai-making-languages-more-accessible-digital-technologyTitle: ONDC Official Seller App SDK Repository [3.1, 3.4, 3.10]
https://github.com/ONDC-Official/seller-app-sdkTitle: BRIA RMBG-1.4 Open-Source Background Removal Model [1.2, 1.5, 2.1]
https://huggingface.co/briaai/RMBG-1.4What we think about this platform, and how it can help people solve SIH problems.
" 30M+ Google Gemini tokens went into researching, analyzing, and understanding all 233 SIH problem statements to make them easier to explore and solve. "
" The platform makes SIH problem statements much easier to understand. Having everything from the problem details to the procedure in one place. "
" Oh, you got ads? Well, we need to make money too! But, you can use our platform to help you win SIH and take home the prize money. "
Get in touch if you need help bringing your ideas to life, or simply have an interesting idea you'd like to talk through. I'm always open to a conversation and exploring new possibilities together.