Voice AI Calling Agent (OnDial)
Enterprise 24/7 Voice Calling Agent Platform
Yellow Verandah
Bespoke Luxury Home Decor E-Commerce Platform
Guru Frogen
Custom Corporate Web Application
Digital Agency Platform
Next-Gen Interactive Web Application
Enterprise RAG
Data Access Across 150+ Warehouses
Teaching Assistant
Multi-Agent Tutoring Platform
Edge PaliGemma
Real-Time Vision-Language Inference
Forest Surveillance
Wildlife & Perimeter Defense
Video Translation
Invisible Dubbing & Lip Sync AI
SAMPARK
Digital Payment Infrastructure
A production-grade, highly optimized Voice AI calling agent platform (ondial.ai) designed to automate 24/7 inbound and outbound enterprise phone conversations. Rather than using default LiveKit/PipeCat wrappers, we engineered a custom low-latency media pipeline delivering audio-in to audio-out latency under 150ms. Powers human-like conversational agents for lead qualification, appointment scheduling, and customer support with enterprise-grade compliance (SOC 2, HIPAA, GDPR, PCI DSS).
A bespoke luxury home decor e-commerce web platform engineered for Yellow Verandah (`yellowverandah.in`). Built to showcase handcrafted artisan collections with high-conversion UX, supporting 10,000+ SKUs, Razorpay payment gateway integration, real-time inventory management, and automated shipping calculations.
A custom enterprise corporate web application engineered for Guru Frogen (`guroufrogen.com`). Built with modern full-stack web technologies to deliver brand positioning, lead capture pipelines, customer onboarding, and dynamic CMS content delivery.
A next-generation interactive agency web application hosted on Vercel (`agency-chi-ten.vercel.app`). Demonstrates fluid 60fps micro-interactions, dark/light theme switching, and optimized component architecture.
Multi-stage advanced RAG pipeline with HyDE, recursive retrieval, schema-aware chunking, graph-based navigation, hybrid search + cross-encoder reranking, and chain-of-table reasoning — deployed across real enterprise supply chain infrastructure.
Two self-hosted LLMs running simultaneously. The visible speech-to-speech tutor interacts live with students. The hidden text-to-text orchestrator silently injects context, memory, difficulty signals, and guidance into the tutor's prompts in real time.
A complete four-stage optimization pipeline taking Google's PaliGemma vision-language model and making it production-ready on Jetson Orin Nano, Raspberry Pi, Android, and iOS — with -65% size reduction and <2% accuracy loss.
Multi-model computer vision ensemble processing drone footage to detect and geolocate perimeter threats in real time.
Fine-tuned diffusion-based lip sync model that translates, synthesizes, and synchronizes lip movements in real time across live video streams.
A full-fidelity simulation platform replicating the multi-party architecture of modern payment ecosystems (10B+ transactions/month capability).