Raju Shingadiya

On-device photo quality checker scoring blur, dark, bright, glare, noise and finger with per-issue scores and hints; tiny, dependency-free, offline inference that ignores screenshots as good.

#web, #wasm, #sdk, #resources, #macos, #kotlin-native, #image, #desktop, #compose-multiplatform, #ai

On-device emoji suggestions from text: returns ranked emoji with names, confidences and optional skin-tone variants; multilingual, private offline inference with tiny footprint and millisecond latency.

#web, #wasm, #text, #sdk, #resources, #native, #kotlin-native, #kotlin-coroutines, #desktop, #apple

On-device personal information detector tailored for Indian formats; identifies and masks 19 PII types (Aadhaar, PAN, UPI, names, addresses) offline, small (8MB), no dependencies.

#web, #wasm, #text, #sdk, #resources, #native, #kotlin-native, #kotlin-coroutines, #compose-multiplatform, #ai

On-device toxic-message detection for chat, handling Indian languages, Romanised scripts and English; fast, small (≈4MB), offline, per-message scores, category hints and adjustable sensitivity.

#web, #wasm, #text, #sdk, #resources, #kotlin-native, #kotlin-coroutines, #desktop, #apple, #ai

On-device English message emotion detection: returns 28 labeled emotions (multiple per message), mood and Ekman basics; tiny offline model (~6.4 MB), fast and privacy-preserving.

#web, #wasm, #text, #sdk, #resources, #native, #kotlin-native, #kotlin-coroutines, #compose-multiplatform, #ai

On-device punctuation and capitalization for chat and voice transcripts, tailored to Indian languages and scripts; uses a tiny transformer, script-specific marks, private offline inference, fast and dependency-free.

#web, #wasm, #text, #sdk, #resources, #kotlin-native, #kotlin-coroutines, #desktop, #compose-multiplatform, #apple

On-device smart-reply suggestions from a curated set of ~1,300 canned replies; private, dependency-free inference, very low latency, one-per-intent chips, configurable confidence and result limits.

#web, #wasm, #text, #resources, #native, #kotlin-native, #kotlin-coroutines, #desktop, #apple, #ai

On-device doodle recogniser reading pen strokes (not pixels), guessing live among 345 QuickDraw categories with a tiny (~1.7 MB) int8 model—offline, dependency-free, low-latency.

#web, #wasm, #sdk, #resources, #performance-optimization, #kotlin-multiplatform-sample, #desktop, #compose-ui, #compose-multiplatform, #ai

On-device language and script detection for 211 labels, including romanised South Asian languages; offline, lightweight (~8.5 MB), dependency-free, fast and accurate on short texts.

#web, #wasm, #text, #sdk, #resources, #kotlin-native, #kotlin-coroutines, #desktop, #compose-multiplatform, #compose