Harnoor Minhas
HM
Harnoor Minhas

AI & Full-Stack Engineer

Agentic AI · RAG · LLM platforms · serverless AWS

Production applications spanning Retrieval-Augmented Generation, LLM pipelines, autonomous agents, real-time voice, and serverless AWS — designed, built, and operated end to end.

AI Voice & Conversational

Innerverse — Voice Companion

Real-time conversational voice app: speech in, LLM reasoning, natural voice out — streamed over WebSocket with sub-second turn-taking.

Claude (Bedrock)ElevenLabsWebSocket APIWeb Audio
Live →

Themeable AI Audio Player

Multi-theme web player with an audio-reactive visualizer (5 themes) driven by live frequency analysis; plays AI-generated two-host "podcast" briefings.

Web Audio APICanvasVanilla JS
Live →

Generative Content Studios

PRISM — Studio Directory

A multiverse hub linking 40+ generative apps across 7 studios, each producing nightly AI content (text, image, audio, video).

CloudFrontS3Multi-tenant
Live →

Manifestation & Reflection Apps

Image + audio generation pipelines (Imagen, Veo, TTS) producing cinematic daily sessions, cached idempotently on S3.

Gemini / Imagen 4Veo 3Lambda
Live →

Daily-Growth App Suite

A studio of habit/growth web apps with nightly LLM-generated content via EventBridge-scheduled Lambdas.

EventBridgeBedrockServerless
Live →

Tools, SaaS & Automation

SnapDomain

AI domain-name generator with live availability checks against Route 53 and full Stripe-powered subscription checkout.

StripeRoute 53 APILambda
Live →

RAG Knowledge Search

Server-side semantic retrieval over a 1,200+ document index: Titan embeddings, cosine ranking, snippet extraction — built to scale past browser-side search.

Titan Embed v2RAGAPI Gateway
Backend service

Automated Newsletter Engine

Research → LLM dedup/extraction → templated render → two-host AI audio → email delivery, with a content-quality CMS and audit trail.

BedrockGemini TTSSES
Listen →

Technical Stack

AI / ML: Amazon Bedrock (Claude Sonnet 4.5 with prompt caching; Titan Embed Text v2, 1024-dim), Google Gemini 2.5 (Pro/Flash/TTS, Imagen 4, Veo 3), ElevenLabs, OpenAI (image, Whisper), Perplexity.

Patterns: Retrieval-Augmented Generation, multi-agent orchestration, structured/JSON LLM output, multi-model fallback & verification, prompt caching, idempotent generation, tiered memory, content-quality gates.

AWS: Lambda, API Gateway (HTTP + WebSocket), DynamoDB, S3, CloudFront, Route 53, ACM, SES, EventBridge, Cost Explorer.

Engineering: Python, JavaScript/Node.js, SQL, HTML/CSS; serverless & event-driven architecture; real-time streaming; Web Audio API; SQLite (vector cache, dedup ledger); Stripe & Twilio integrations.