YoAdvait / Loading
Loading pageIndex / Projects
Things I’ve built across AI, audio, and software. Mostly solving real problems for clients or products of my own.
Founder-built AI governance layer for Indian NBFC lending, real-time explainability, adaptive threshold drift detection, and audit-trail infrastructure for RBI-aligned credit decisions.
AI system that automates content classification and metadata enrichment for a music tech client. CNN-LSTM hybrid achieves 89% accuracy across 30+ genres/classes for genre classification, enabling the platform to tag and categorize music at scale without manual review.
Real-time visual performance software by SummonSound for VJs, DJs, and live productions. Combines multi-layer video playback, projection mapping, audio-reactive effects, MIDI control, and DMX lighting into a low-latency performance environment built for live shows.
Audio intelligence engine developed for TuneM that matches vocals with instrumentals and generates personalized recommendations. It measures compatibility across timbral, rhythmic, harmonic, and metadata features, then ranks the closest-fitting tracks for each vocal.
AI-powered VST/AU plugin for game audio and film production that detects and isolates footstep sounds in real time using neural network inference, automating a task that would otherwise require manual audio editing.
Local-first AI production system that turns scripts into finished narration and video. Combines multi-model text-to-speech, timing optimization, audio mastering, AI-assisted visual generation, animation, and timeline editing into a single production pipeline.
Commercial MIDI plugin developed for a client around their proprietary interaction concept. I built the real-time gesture-processing and arpeggiator system that translates expressive X-Y touch movement into playable MIDI patterns, with gesture analysis, dynamic sequencing, parameter mapping, quantization, and low-latency MIDI output.
Professional desktop application that automates music library organization: intelligent tagging, metadata editing, and batch processing, reducing hours of manual work to minutes for music producers and labels.
Commercial virtual instrument developed for 12 Bit Soul, built around a custom multi-sample playback engine with velocity layers, polyphonic voice handling, MIDI performance controls, sound-shaping DSP, and a complete preset system. Delivered as production-ready VST3 and AU builds for macOS and Windows.
Full booking platform for a premium mountain homestay: automated inquiry pipeline, Supabase backend, and dual email notifications. The minimalist design communicates exclusivity while the backend eliminates manual booking overhead.