KalpLathia

Full-StackAIArchitect

Scroll

01 / About

The Engineer

About Me

Senior Full-Stack Engineer with 6+ years of experience building scalable web platforms and production-grade AI systems. Specialized in architecting end-to-end applications combining React/Next.js, Node.js/FastAPI, and multi-agent LLM workflows. I own system architecture from frontend to infrastructure, delivering robust, high-performance applications.

My journey started with Angular frontends for healthcare clients, evolved through blockchain marketplaces and AI-powered platforms, and led me to where I am now — architecting multi-agent systems that plan, reason, and execute autonomously.

From pixel-perfect interfaces to autonomous agent systems — a journey driven by curiosity and craft.

Let's Connect

Resume

Years Exp

Companies

Github Contribs

Technologies

AI Products

Automations

Command Centers

Blockchain

Websites

02 / Journey

The Evolution

Eighteen products across four chapters. Each chapter marks a paradigm shift in how I think about building software.

Chapter I

The Interface Crafter

Building pixel-perfect experiences for the web

“Every journey starts at the edge of the screen — where code meets the people who use it.”

Ch. I·Frontend EngineerWebsite

Hyperactive

Healthcare intelligence on every tablet

An iPad application for a US-based healthcare company with a bridge between AngularJS and native iOS. Hospital search with multi-parameter filtering, heavy price calculations across treatments, analytics and reporting. Generated PDF reports up to 45–50 pages and Excel reports.

Where I learned that good interfaces aren't about aesthetics — they're about reducing cognitive load when the stakes are high.

•Multi-parameter hospital search and filtering
•Heavy price calculations across treatments
•PDF reports up to 45–50 pages

AngularJSTypeScriptiOS Native BridgePDF GenerationExcel ReportsREST APIs

Ch. I·Frontend EngineerWebsite

TKE (TK Elevator)

Dynamic enterprise forms at scale

Enterprise project for TK Elevator via Infosys. Built dynamic Angular forms where the entire structure came from the backend. Handled heavy calculations and complex logical formulas. NX workspace with shared libraries across multiple Angular projects.

Enterprise-scale frontend — where every form is dynamic and every calculation has to be exact.

•Dynamic form rendering from backend config
•Complex calculation logic and formulas
•NX monorepo with shared libraries

AngularTypeScriptNX WorkspaceNX LibrariesDynamic FormsREST APIs

Ch. I·Frontend EngineerWebsite

4Bit Technology

A developer-first company presence

Company website for 4Bit Technology featuring a dark-themed design with a static chatbot widget for visitor engagement.

Clean, minimal — sometimes the best interfaces are the ones that get out of the way.

•Dark-themed responsive design
•Static chatbot widget integration
•Performance-optimized static site

AngularTypeScriptHTMLSCSS

Ch. I·Frontend EngineerWebsite

Journey2Germany

Pixel-perfect from Figma to production

Angular website for a Germany immigration consultancy. SSR-enabled for SEO. Integrated Calendly for scheduling and Google Search Console. Optimized for Lighthouse scores.

The project where pixel-perfect stopped being a buzzword and became a practice.

•Pixel-perfect implementation from Figma
•SSR-enabled for SEO performance
•Optimized Lighthouse scores

AngularTypeScriptAngular Universal (SSR)CalendlyGoogle Search ConsoleSCSS

“Interfaces taught me empathy for users. But I wanted to build systems that didn't require trust — they proved it.”

Chapter II

The Trust Architect

Building verifiable systems on the blockchain

Ch. II·Frontend LeadBlockchain

NFT Marketplace

The eco-friendly bazaar of the blockchain

An eco-friendly NFT marketplace on the Algorand blockchain. Users mint, sell, buy, rent, and resell NFTs. Featured single-step minting, partial minting, and a White Label NFT solution (NFTaaS). Supported credit card and crypto payments. Artists receive 90% from initial sales and 5% royalties.

My first deep dive into decentralized systems. Single-handedly built the entire frontend while learning to think in transactions instead of API calls.

•Single-handedly built the complete frontend
•Web3 wallet authentication and real-time transactions
•White Label NFT solution for enterprises

AngularTypeScriptPrimeNGWeb3Algorand BlockchainMetaMask WalletREST APIs

Ch. II·Frontend LeadBlockchain

LicitDoc

Trust, verified on the blockchain

A Web3 digital asset verification platform on the Algorand blockchain. Institutions upload student documents which get a permanent metahash on-chain. Features drag-and-drop verification, near-instant authentication, built-in OCR AI, and counterfeit detection. Launched at Algorand Decipher 2022 in Dubai.

Where I learned that the best systems are invisible. The user never thinks about the blockchain underneath — they just trust the document.

•Blockchain-backed credential verification
•OCR AI for document classification
•Launched at Algorand Decipher 2022

AngularTypeScriptAlgorand BlockchainOCR/AISecure Document StorageREST APIs

“Blockchain showed me the power of trustless systems. Building complete products meant owning every layer — from database to deployment.”

Chapter III

The Full-Stack Builder

Owning the stack from database to deployment

Ch. III·Full-Stack EngineerCommand Center

Kwickprep

40+ modules for education management

A comprehensive education management platform with 40+ modules — academic management, exams, fees, attendance, users, batches, schedules, homework, discussion forums, and more. Integrated WhatsApp Business API and an AI-powered Study Buddy.

The biggest dashboard I've built. 40+ modules taught me that scale isn't about complexity — it's about consistency.

•40+ modules spanning full institutional lifecycle
•AI-powered Study Buddy for students
•WhatsApp Business API integration

AngularTypeScriptPrimeNGNode.jsMongoDBOpenAI APIsWhatsApp Business APIAWS S3AWS EC2

Ch. III·Full-Stack EngineerCommand Center

AGS (All Gift Studio)

AI-powered catalogue to cart

Business operations dashboard for a corporate gifting company. AI-powered Product Extractor — upload a catalogue PDF, AI extracts products with image coordinates for bulk creation. AI product search narrows from themes to categories to price to 5 curated results. Entire flow automated via WhatsApp.

The product where AI stopped being a feature and became the workflow. From PDF upload to product creation — zero manual entry.

•AI-powered bulk product extraction from PDF catalogues
•Automated product search flow via WhatsApp
•Campaign scheduling and analytics

AngularTypeScriptPrimeNGNode.jsMongoDBAWS BedrockOpenAI APIsWhatsApp Business APIAWS S3GitHub Actions

Ch. III·Full-Stack EngineerCommand Center

J2G Dashboard

Operations hub for education consultancy

Operations dashboard for a consultancy helping students study or work in Germany. Real-time chat with groups, content management for 2000+ courses and 200+ universities, customer enrollment and plan management, payment tracking, and automated sync scripts.

Where I learned to build systems that scale with data — 2000+ courses, 200+ universities, all searchable and synced.

•Real-time chat with group support
•2000+ courses and 200+ universities managed
•Automated backup and cross-platform sync

Next.jsReactTypeScriptMongoDBshadcn/uiTailwind CSSFirecrawlAWS S3Google Cloud RunGitHub Actions

Ch. III·Full-Stack EngineerAutomation

MV Maverick

From an email to a complete agent profile

Automation platform for real estate agent data enrichment and marketing video creation. Enriches agent profiles from just an email address. Integrated JSONtoVideo for building marketing videos with customizable scenes, animated slides, and intro slides.

The bridge between AI and automation — replacing hours of manual research with seconds of intelligent scraping.

•Agent data enrichment from email input
•Marketing video creation via JSONtoVideo
•Multi-source web scraping pipeline

Next.jsReactPythonSeleniumWeb ScrapingJSONtoVideo APIMongoDBREST APIs

Ch. III·Full-Stack EngineerAutomation

Email Automation

Intelligent email management at scale

Connect any email account and manage emails intelligently. AI analyzes content to generate summaries, group threads, and organize into labeled collections. Smart categorization, priority inbox, bulk actions, and scheduled composition.

Building an AI layer on top of email — turning inbox chaos into organized, actionable intelligence.

•AI-powered email summarization and grouping
•Smart categorization with auto-labeling
•Real-time sync via WebSockets

Next.jsReactTypeScriptMongoDBOpenAI APIsshadcn/uiTailwind CSSWebSocketsOAuth 2.0IMAP/SMTP

Ch. III·Full-Stack EngineerAutomation

WhatsApp Automation

Custom WhatsApp workflows, zero limits

WhatsApp Business API automation platform with a fully custom chat UI. Template messaging, campaign scheduling for bulk messaging, customizable conversation flows, configurable auto-replies, and AI-powered smart replies with intent detection.

When the platform gives you no UI, you build one — and make it better than the original.

•Fully custom chat UI for WhatsApp Business
•Campaign scheduling for automated bulk messaging
•AI-powered smart replies and intent detection

Next.jsReactTypeScriptMongoDBWhatsApp Business APIOpenAI APIsshadcn/uiWebSocketsWebhooks

Ch. III·Full-Stack EngineerAutomation

Web Scraping Automation

Structured intelligence from any website

Structured web scraping pipeline — takes URLs, scrapes in markdown and HTML, chunks by content size, passes to LLM for structured output. Master file comparison generates diff reports. Historical storage and vector DB for semantic retrieval.

Turning unstructured web content into clean, queryable data — with version tracking and semantic search.

•LLM-powered structured data extraction
•Master/diff comparison across scrape iterations
•Vector DB for semantic query retrieval

PythonSeleniumPlaywrightMongoDBOpenAI APIsPineconeREST APIs

“Full-stack gave me the power to build anything. Adding intelligence to those systems — that changed everything.”

Chapter IV

The Intelligence Architect

Engineering intelligence into every layer

Ch. IV·Full-Stack AI ArchitectAI Product

AIChefs

Where recipes meet artificial intelligence

Users describe their cravings and AI generates custom recipes with photos. Social features — bookmark, like, share, follow. Integrated Instacart for direct ingredient ordering. SSR-enabled with WebSocket streaming. Led and mentored 2 junior developers.

Our first AI product. Where I discovered the magic of LLMs — turning vague human desire into something tangible and actionable.

•AI-powered recipe and image generation
•Instacart integration for ingredient shopping
•SSR-enabled with real-time streaming

AngularTypeScriptPythonOpenAI APIsWebSocketsSSRMongoDBInstacart APIREST APIs

Ch. IV·Full-Stack AI ArchitectAI Product

Uprate.AI

Intelligence embedded in every storefront

AI chatbot evolved into a Shopify app. Pre-trained AI characters match merchant brand tone. Learns product catalog in under a minute. Customers discover products, get answers, and add to cart from chat. 5 stars, 64 reviews on Shopify App Store.

The product that taught me about platform evolution. From standalone chatbot to Shopify app — each pivot sharpened the product instinct.

•Pre-trained AI characters matching brand tone
•Product catalog learned in under 1 minute
•5 stars with 64 reviews on Shopify

AngularTypeScriptPythonRemixOpenAI APIsLlamaIndexMongoDBShopify APIsREST APIs

Ch. IV·Full-Stack AI ArchitectAI Product

Closearch

AI eyes on every property listing

Real estate AI platform with natural language property search. Images analyzed via OpenAI Vision, embeddings stored in Pinecone. Search by visual features — "property with pool in backyard". Map view, collections, bookmarks. Agent tools for document analysis and contract building.

The most technically ambitious project. Combining vector databases, computer vision, and LLMs into a single search experience that felt like magic.

•Vision API-powered image analysis and vector search
•Natural language property search with confidence scoring
•Document analysis and contract builder for agents

Next.jsReactPythonOpenAI APIsPineconeMongoDBMLS APIsGoogle Maps APIREST APIs

Ch. IV·Full-Stack AI ArchitectAutomation

Browser Automation

Multi-agent browser control with vision

Multi-agent browser automation system. Planner agent breaks queries into steps, orchestrator delegates to specialized sub-agents — Navigation, Interaction, Form, Wait, Validation. After every tool call, captures DOM snapshot with highlighted elements and screenshot for vision-based state awareness.

Where I truly understood tool-based AI — an orchestrator directing specialized agents, each with their own LLM client and tools.

•5 specialized sub-agents with own LLM clients
•Vision-based DOM state awareness after every action
•Full browser context management across tabs

PythonPlaywrightOpenAI APIs (GPT-4 Vision)Multi-agent ArchitectureTool Orchestration

Ch. IV·Lead Engineer & ArchitectAI Product

RA-Core

One system to orchestrate them all

Unified multi-agent framework consolidating all agent capabilities. Automatically analyzes queries, creates execution plans, selects tools and agents, executes, and delivers results via email. Supports web search, document processing, image processing, verification, and reporting.

The culmination of everything. A single system that plans, delegates, and executes — the architecture I'd been building toward since my first Angular component.