Work.

Freya

Real-Time AI Voice Assistant & Receptionist

Built a production-grade AI receptionist for natural, real-time customer conversations. Freya answers questions with Retrieval-Augmented Generation (RAG), performs backend actions, schedules appointments, automates business workflows, and transfers conversations to a human operator when confidence is low. Powered by LiveKit, Deepgram, Cartesia, Groq-hosted LLMs, Hugging Face Embeddings, and Upstash Vector, the system preserves conversational context, retrieves verified answers from a business knowledge base, and uses tools to complete tasks—not merely generate responses. The architecture separates real-time voice transport from agent orchestration, keeping latency low while supporting multi-turn conversations, tool execution, retries, and future expansion to web and messaging channels. This separation also makes the system easier to test, maintain, and extend.

Node.jsExpress.jsJavaScriptLiveKit Agents FrameworkOpenAI APIOpenAI EmbeddingsOpenAI TTSDeepgram APISpeech-to-TextText-to-SpeechUpstash Vector DatabaseVector EmbeddingsPostgreSQLReal-time CommunicationWebRTCVoice Activity Detection (VAD)Silero VADBackground Noise CancellationNatural Language ProcessingSemantic SearchAPI IntegrationSession ManagementEvent-Driven ArchitectureError HandlingDatabase DesignModular ArchitectureTool DevelopmentConversational AIVoice Agent Development

link:

IndStudio AI Trialroom

Published Generative AI Shopify App

Architected and launched IndStudio AI Trialroom, a published Shopify app that lets shoppers upload a photo and visualize themselves wearing a store's apparel through generative AI. The product delivers a complete merchant experience with an embedded admin dashboard, analytics, billing, product management, and storefront theme extensions. An asynchronous backend built with AWS SQS, Redis, PostgreSQL, and Server-Sent Events handles long-running AI jobs while streaming progress updates to users in real time. Each generated image is evaluated for identity preservation, garment fidelity, pose consistency, and overall visual quality before delivery. Owned the platform end to end—from AI integration and cloud infrastructure to Shopify APIs, App Store compliance, GDPR webhooks, subscriptions, analytics, and deployment—turning a complex AI workflow into a reliable commerce product.

TypeScriptRemixShopify App DevelopmentAI Image GenerationPrompt EngineeringPostgreSQLRedisAWS(S3 & SQS)

link:

Aiversety

RAG-Powered Learning & Knowledge Platform

Built Aiversety, an AI learning platform that turns PDFs and web content into interactive, searchable knowledge bases. Users can ask questions about their material, generate quizzes and topic summaries, and receive grounded answers with source citations. Designed as a scalable SaaS product, Aiversety combines LangChain, semantic vector search, structured AI generation, authentication, asynchronous document processing, and knowledge management to support dependable learning workflows beyond a basic document chatbot. Content is parsed, chunked, embedded, and indexed before retrieval through a RAG pipeline. Schema-based outputs keep generated quizzes and summaries consistent and ready for use in the application. The project covers the complete product lifecycle, including system architecture, frontend development, retrieval pipelines, cloud deployment, and scalable backend infrastructure.

TypeScriptNext.jsPrompt EngineeringLLM IntegrationAuthenticationStructured Output ParsingVector Store IntegrationRAGSemantic SearchPostgreSQLRedis

link:

Services.

About.

I build production-ready AI agents, LLM integrations, automation systems, and SaaS products that solve real business problems. My work includes a published AI Shopify app, real-time voice assistants, RAG-powered knowledge platforms, and enterprise automations used in daily operations. I focus on building AI systems that go beyond generating responses—they retrieve reliable information, execute business actions, automate workflows, and collaborate with human operators when needed. I take ownership from architecture through deployment, work through ambiguity, and turn complex requirements into dependable products. AI accelerates my development process, but every solution is reviewed, tested, secured, and structured for maintainability after handoff.