# Hamza Kahloon — Full Stack Developer & AI/ML Engineer

> Hamza Kahloon is a Lahore-based Full Stack Developer and AI/ML Engineer with 4+ years of experience building AI voice agents, RAG pipelines and multi-tenant SaaS platforms with React, Next.js, Node.js, Python and FastAPI.

## At a glance

- Name: Hamza Kahloon (full name: Muhammad Hamza Tanveer Kahloon)
- Role: Full Stack Developer & AI/ML Engineer (Full Stack Developer, AI / ML Engineer, Voice AI Specialist, SaaS Architect)
- Based in: Lahore, Pakistan · works with teams worldwide
- Current company: Decoders Digital
- Experience: 4+ years building, 25+ projects shipped, 15+ ai models deployed, 1,000+ daily active users
- Education: BS Software Engineering, The University of Lahore (2022, CGPA 3.3 / 4.0)
- Availability: Open to full-time roles & freelance projects
- Website: https://hamzakahloon.online
- Email: hamzaakahloon903@gmail.com
- Phone / WhatsApp: +92 309 1453950 (https://wa.me/923091453950)
- LinkedIn: https://www.linkedin.com/in/hamza-kahloon-12a14125a/
- GitHub: https://github.com/Hamza-Kahloon786

## About

I'm Hamza — a full stack developer who specialises in AI. For more than four years I've been turning ideas into intelligent products: voice agents that pick up real phone calls, multi-tenant platforms that clinics and field-service teams run on daily, and retrieval systems that make documents genuinely useful.
I like owning the whole lifecycle — the interface, the API, the model integration and the deployment — because that's where a product actually comes together. Every project here was built to serve real users, not to sit in a demo.

### What he focuses on

- **AI Engineering:** LLMs, RAG pipelines, LangChain & LangGraph agents, NLP, OCR and evaluation frameworks that prove the model works.
- **Full-stack Product:** React, Next.js and Angular front ends on Node.js, FastAPI and Django back ends — multi-tenant, secure and fast.
- **Voice & Automation:** Twilio, ElevenLabs, Vapi, Whisper and Deepgram voice agents, wired into n8n workflows across voice, SMS, email and chat.

## Experience

### Full Stack Developer — AI/ML — Decoders Digital (2025 — Present)

Full-time · Lahore, Punjab. Building AI-first SaaS products end-to-end — voice agents, healthcare platforms and ML systems — and running them in production.

- Developed LeadFront (formerly Vendira.ai), a full-stack AI voice agents SaaS automating phone calls, customer support and sales across voice, email, SMS and chat (Twilio, Deepgram, ElevenLabs, OpenAI, n8n).
- Led the SaaS transformation of Doctor AI Portal into a multi-tenant, HIPAA-focused healthcare platform with tenant-level data isolation, role-based access control and AI-powered patient intake (Whisper / Deepgram).
- Built an AI/ML bird sound recognition system using audio signal processing and machine learning for automated species classification.
- Deployed and maintained production applications with Docker, Gzip compression and traffic handling for high-load environments.
- Stack: React, Node.js, PostgreSQL, Twilio, ElevenLabs, OpenAI, n8n, Whisper, Deepgram, Docker

### Full Stack Developer — AI/ML — Pangea Global Enterprise (PGE) (2022 — 2025)

Full-time · 3 years · Lahore, Punjab. Architected enterprise AI/ML web applications serving multiple organisations on a multi-tenant architecture.

- Architected and deployed enterprise-grade AI/ML web applications using React.js, Next.js, Angular, Node.js and Python, serving multiple organisations with multi-tenant architecture.
- Developed and integrated machine learning models for intelligent data processing, automated decision-making and real-time conversational AI systems.
- Built robust REST APIs and FastAPI microservices for frontend–backend communication and AI model deployment, documented with OpenAPI.
- Implemented NLP-powered chatbots, voice AI (Vapi, Twilio, ElevenLabs), RAG systems and LangChain / LangGraph agent workflows.
- Engineered real-time communication with WebSockets, voice/SMS capabilities and multi-channel automation using n8n workflows.
- Optimised production deployments on VPS, Vercel and Railway with Docker, Gzip compression and traffic handling for high-load applications.
- Stack: React, Next.js, Angular, Node.js, Python, FastAPI, LangChain, LangGraph, Vapi, WebSockets, Docker

### Software Developer — Freelance (2021 — 2022)

Freelance · Lahore · Remote. Where it started — building software for real clients while studying Software Engineering.

- Built Java applications for clients, including a complete hospital management system.
- Learned to own delivery end-to-end: gathering requirements, building, shipping and supporting real users.
- Stack: Java, MySQL, HTML5, CSS, JavaScript

## Projects

### LeadFront — The AI front office for service businesses

- Slug: leadfront
- Year: 2025 — 2026 · Role: Full Stack Developer · Decoders Digital · Status: Production
- Categories: Voice AI, SaaS, AI & LLM
- Stack: React, Vite, Node.js, Twilio, Deepgram, ElevenLabs, OpenAI, RAG, Stripe, Google Calendar, n8n
- Results: 500+ businesses automated; 2M+ minutes handled; 10+ languages, auto-detected
- Links: [Case study](https://hamzakahloon.online/projects/leadfront) · [Live site](https://leadfront.net)

An AI business-communication platform: voice and SMS agents that answer every call, qualify leads, book appointments and follow up automatically — with a built-in CRM, campaigns, job scheduling and a technician portal.

LeadFront (formerly Vendira.ai) clones a business's best rep. AI voice and SMS agents answer instantly, qualify leads, book appointments and follow up — 24/7, in 10+ languages, on the business's existing phone number.
I built the platform end-to-end at Decoders Digital: the React app, the Node.js services, the real-time voice pipeline and the automations that tie calls, SMS, email, CRM and scheduling into one system.

What was built:
- AI inbound receptionist and outbound call campaigns with human-like voices — unlimited parallel calls and ~2-second average response.
- Real-time voice pipeline: Twilio telephony, Deepgram speech recognition, OpenAI conversational intelligence and ElevenLabs voices.
- Agents auto-detect and switch between 10+ languages mid-call, including Spanish, Hindi and Urdu.
- AI agent builder with custom scripts, knowledge-base uploads, premium voice selection and RAG-powered contextual answers.
- Appointment booking during live calls with Google Calendar sync and automated SMS reminders.
- SMS and email campaigns — bulk sends, templates, delivery / open / click tracking, auto-replies and follow-up sequences.
- Unified inbox, built-in CRM with lead scoring and pipeline, a job scheduler and a technician portal for field teams.
- Call analytics with AI summaries, outcome tracking, keyword extraction, transcripts and recording playback.
- Integrations with Salesforce, HubSpot, Jobber, Stripe, Slack, Zapier and Calendly; AES-256 encryption and role-based access.

### Doctor AI Portal — Multi-tenant healthcare SaaS with an AI front desk

- Slug: doctor-ai-portal
- Year: 2025 — 2026 · Role: Lead Full Stack Developer · Decoders Digital · Status: Production
- Categories: Healthcare, SaaS, Voice AI
- Stack: React, Node.js, PostgreSQL, Whisper, Deepgram, OpenAI
- Results: 4 access roles; 2 speech-to-text engines; SOAP notes auto-generated
- Links: [Case study](https://hamzakahloon.online/projects/doctor-ai-portal) · [Live site](https://platform.medaipartners.com)

Transformed a single-clinic doctor portal into a multi-tenant, HIPAA-focused healthcare SaaS with AI voice & chat patient intake and automated SOAP notes.

Doctor AI Portal (MedAI Partners) is an autonomous front desk for clinics — it handles patient intake, refill requests and scheduling, then hands doctors structured notes to review.
I led the transformation from a single-clinic system into a true multi-tenant platform: every query scoped by clinic, and every clinic with its own branding, onboarding and reporting.

What was built:
- Tenant-level data isolation with clinic_id scoping, plus per-clinic branding, onboarding and reporting.
- AI voice and chat intake with Whisper / Deepgram transcription and automated SOAP-note generation with a doctor review and approval workflow.
- Prescription refill requests with approve / deny workflow and e-signature support.
- Appointment scheduling with provider slot management, waitlist auto-promotion and patient notifications.
- Role-based access control across Super Admin, Clinic Admin, Provider and Patient with a permission matrix.
- HIPAA-focused security: encryption at rest and in transit, audit logging, session monitoring and a Super Admin dashboard for clinic, revenue and provider metrics.

### STORM AI — AI-powered CRM & field-service platform

- Slug: storm-ai
- Year: 2025 · Role: Full Stack Developer · Status: Production
- Categories: SaaS, AI & LLM
- Stack: Next.js, React, Supabase, Node.js, FastAPI, MongoDB, Twilio, QuickBooks, Stripe, Google Calendar
- Results: 1,000+ daily active users; <150ms response time; 99.9% uptime
- Links: [Case study](https://hamzakahloon.online/projects/storm-ai) · [Live site](https://stormai.decodersdigital.net) · [Source code](https://github.com/Hamza-Kahloon786/Saas)

An enterprise multi-tenant CRM for service companies — leads, scheduling, dispatch, invoicing and an AI sales assistant that works over SMS.

STORM AI runs the day-to-day of field-service businesses: from the first inbound lead to the technician arriving on site and the invoice being paid.
An AI sales assistant qualifies leads and handles objections over SMS, while a custom AI flow editor lets each company build its own automations.

What was built:
- Multi-tenant architecture with role-based access control.
- Contact & lead management, drag-and-drop job scheduling, route optimisation and GPS tracking.
- AI Sales Assistant with SMS conversations, lead qualification and objection handling.
- Automation workflows with Twilio, Google Calendar, QuickBooks and a custom AI flow editor.
- Customer portal plus an analytics dashboard with KPIs, lead-conversion tracking and document management.

### AIPAS — AI procurement automation & RAG system

- Slug: aipas
- Year: 2023 — 2024 · Role: Full Stack & AI Engineer · Status: Live
- Categories: AI & LLM, Machine Learning
- Stack: Python, Django, FastAPI, React, MongoDB, OCR, NLP, RAG, Chroma, FAISS, Pinecone, Gradio, Hugging Face
- Results: 92% decision accuracy; 1,000+ documents / hour; 35% cost reduction
- Links: [Case study](https://hamzakahloon.online/projects/aipas) · [Live site](http://aipas.duckdns.org) · [Source code](https://github.com/Hamza-Kahloon786/Procurement-Automation-System) · [Demo video](https://www.linkedin.com/posts/hamza-kahloon-12a14125a_project-title-aipas-ai-powered-procurement-ugcPost-7458551299677118464-T92U/)

Converts email-based procurement requests into structured, automated decisions with OCR, NLP and ML — plus a full RAG evaluation framework.

AIPAS — live as ProcureHub — connects buyers and vendors on one platform and automates the slow parts of procurement: reading documents, extracting requirements, comparing quotations and choosing the right vendor.
On top sits a RAG comparison system that benchmarks vector stores and retrieval strategies, so every answer the assistant gives can be measured.

What was built:
- OCR + NLP pipeline for automated document processing and intelligent vendor selection.
- Separate buyer, vendor and admin portals — buyers raise requests and track open requests, quotations received and completed deals.
- AI-driven quotation analysis and an in-app AI chat assistant for procurement questions.
- 92% accuracy on automated procurement decisions through ML algorithms.
- RAG comparison system deployed on Hugging Face Spaces with a Gradio UI and an MCP server.
- Hierarchical metadata classification with a 3-level taxonomy (domain / section / topic).
- Vector search across Chroma, FAISS and Pinecone with metadata filtering.
- Evaluation framework with Hit@k, MRR, semantic similarity and latency metrics; presets for hospital, bank and fluid-simulation domains.

### EZOS — Proposals, commissions & payroll for sales teams

- Slug: ezos-proposal-tool
- Year: 2026 · Role: Full Stack Developer · Status: Production
- Categories: SaaS
- Stack: Next.js, React
- Results: 6 commission tiers per sale; 3 brands supported; 0 deploys to change pricing
- Links: [Case study](https://hamzakahloon.online/projects/ezos-proposal-tool) · [Live site](https://ezos.online)

A sales proposal engine and commission tracker for water-treatment sales teams — guided proposals with enforced pricing, branded PDFs, multi-level rep commissions and payroll reporting.

EZOS takes a sales team from the first conversation to a signed proposal: reps build a proposal in minutes, and every document goes out as the same branded PDF.
Behind it sits a commission engine that pays out across six tiers of the sales hierarchy, plus a no-code control panel so admins change packages, pricing and financing without a deploy.

What was built:
- Guided proposal builder — customer, package, upgrades, margin and financing in one flow.
- Admin-set margin ceilings and financing terms, enforced on every proposal.
- Branded PDF proposals that print, download and email identically.
- Commission dashboard: total sales, revenue and commission owed, revenue and sales by package, and commission by rep with date-range filters.
- Sales ledger filtered by rep, water type and status, with install and payment tracking and PDF export.
- Multi-level commission splits per package — rep, direct recruiter, team lead, regional, partner and override.
- No-code control panel for packages, product types, water types, offices, financiers and adders — changes take effect immediately.
- Payroll reporting and user management for authorised staff across the Supreme, Homewater and H2Pros brands.

### ChapterLens — AI that keeps technical books up to date

- Slug: chapterlens
- Year: 2026 · Role: Full Stack & AI Engineer · Status: Live
- Categories: AI & LLM, SaaS
- Stack: Next.js, React, OpenAI
- Results: 14 pipeline stages; 4 export formats; 7+ focus areas
- Links: [Case study](https://hamzakahloon.online/projects/chapterlens) · [Live site](https://chapteragent.spacetechnologyseries.com)

Upload a technical book chapter and ChapterLens finds outdated facts, researches current sources, proposes style-matched rewrites with citations and exports a tracked, print-ready document.

Technical books go out of date fast. ChapterLens scans every factual claim in a chapter, flags what's stale and researches what's true today.
Authors stay in control: every proposed change comes with cited sources, can be approved, edited or rejected, and is rewritten to match their own voice.

What was built:
- Drag-and-drop DOCX upload up to 50 MB — text, figures, tables and equations extracted instantly.
- GPT-4o scans every factual claim, flags outdated content and profiles the author's writing style.
- Automated web research pulls current data from government, academic and technical sources, cited per claim.
- Side-by-side review of original vs. updated text — approve, edit or reject each change.
- Style-matched rewrites that keep the author's tone, grade level and sentence structure.
- Exports an updated DOCX, a highlighted DOCX, PDF previews, a CSV changelog and a full audit trail.

### ChainScope AI — Carbon intelligence for UK supply chains

- Slug: chainscope-ai
- Year: 2026 · Role: Full Stack & ML Engineer · Status: Open source
- Categories: AI & LLM, Machine Learning, SaaS
- Stack: React, Vite, Tailwind CSS, FastAPI, MongoDB, scikit-learn, OpenAI, Google Maps, Stripe
- Results: 12-mo emission forecasts; Scope 1–3 emissions tracked; 48h grid forecast
- Links: [Case study](https://hamzakahloon.online/projects/chainscope-ai) · [Source code](https://github.com/Hamza-Kahloon786/Echochain)

A full-stack SaaS that finds carbon hotspots across UK supply chains, forecasts emissions with ML and recommends reductions on an interactive map.

ChainScope AI (the app ships as EchoChain — Carbon Identifier) gives sustainability teams one view of their whole supply chain — suppliers, warehouses and transport routes — with the emissions of each one calculated automatically.
Machine-learning forecasts and an AI assistant turn that data into actions: which routes to switch, which sites to fix first, and a compliance-ready report.

What was built:
- Scope 1 / 2 / 3 tracking using UK DEFRA 2024 emission factors for fuel, electricity, gas and freight — with a dashboard of total emissions, scope breakdown, monthly trend and emissions by category.
- Carbon hotspot map with suppliers, warehouses and routes — AI-generated insights on every marker.
- RandomForest regression forecasting — historical vs. forecasted emissions with an adjustable horizon of up to 12 months.
- Route optimisation on real roads via Google Directions with per-mode savings (road, rail, sea, air).
- Live UK grid page from the National Grid ESO Carbon Intensity API — current gCO₂/kWh with a rating, renewable / low-carbon / fossil share, generation mix and a 48-hour forecast.
- AI Assistant on GPT-4o-mini — risk-rated recommendations with quick wins and prioritised, scope-tagged actions, an Ask AI chat and SECR compliance report generation.
- Stripe subscriptions and Excel / CSV bulk import for suppliers.

### Legal Assistant AI — Enterprise RAG for legal research

- Slug: legal-assistant-ai
- Year: 2025 · Role: Full Stack & AI Engineer · Status: Case study
- Categories: AI & LLM
- Stack: React, Node.js, OpenAI, Chroma, RAG
- Results: 94% answer accuracy; 70% faster research; 5+ law firms
- Links: [Case study](https://hamzakahloon.online/projects/legal-assistant-ai) · [LinkedIn write-up](https://www.linkedin.com/posts/hamza-kahloon-12a14125a_ai-legaltech-fullstackdevelopment-activity-7372317966253645825-dIFH)

An enterprise RAG-powered legal assistant that retrieves the right clauses and precedents from a vector store and answers with GPT-4 — deployed at 5+ law firms.

Lawyers ask questions in plain language; the assistant embeds the query, retrieves the most relevant passages from a ChromaDB vector store and has GPT-4 answer strictly from that context.
The result is research that is faster and traceable — every answer comes with the sources it was built from.

What was built:
- Retrieval pipeline: embedding model → ChromaDB vector store → top-k retrieval → GPT-4 with grounded context.
- Answers cite the exact section and source file (e.g. Section 148(g) of the Legal Profession Act 1966) and show a confidence level plus the top source chunks with similarity scores.
- Document library: upload legal PDFs, automatic chunking (141 chunks for a single Act), processing status, search and type filters.
- Chat history with search, favourites and filters; save and export conversations.
- Built-in legal disclaimers on every response and role-based accounts for lawyers.
- 94% answer accuracy on legal queries.
- 70% faster legal research for the teams using it.
- Deployed at 5+ law firms, saving an estimated $50K+ per firm annually.

### TripAI — AI trip planner with streaming itineraries

- Slug: tripai
- Year: 2026 · Role: Full Stack & AI Engineer · Status: Open source
- Categories: AI & LLM, Machine Learning
- Stack: React, Redux, Node.js, Express, MongoDB, Redis, OpenAI, Flask, scikit-learn, Google Maps
- Links: [Case study](https://hamzakahloon.online/projects/tripai) · [Source code](https://github.com/Hamza-Kahloon786/AI-Based-Trip-Planner) · [Demo video](https://www.linkedin.com/posts/hamza-kahloon-12a14125a_machinelearning-artificialintelligence-mern-ugcPost-7486188742291079168-yAGk/)

Travellers across Pakistan describe their preferences and get a complete, personalised itinerary in minutes — streamed live from GPT-4.1-mini and enriched with real routes, live weather, ML cost predictions and budget-matched hotels.

TripAI turns a short multi-step form into a full day-by-day trip — say Lahore to Hunza Valley or Naran. The itinerary streams in as it's generated, so users watch their plan take shape.
Behind the scenes, AI jobs run on a Redis-backed queue and a separate Python service handles the machine-learning recommendations.

What was built:
- Streaming itineraries over Server-Sent Events from OpenAI GPT-4.1-mini.
- Background AI jobs with Redis + BullMQ workers for reliable generation.
- Python Flask ML microservice — RandomForest regression and content-based filtering.
- Real routes from Google Maps — distance, travel time and alternative routes on an embedded route map.
- Live weather from OpenWeatherMap plus a climate-suitability score per destination and month, with best months to visit and warnings for snow-blocked passes.
- ML cost prediction with a per-person budget breakdown in PKR, and hotel recommendations matched to the budget.
- Saved trip plans for every user, downloadable as PDF.
- React 19 front end with Redux Toolkit, React Query and validated multi-step forms.

### GarageAI — Missed calls → ready-to-book leads for UK garages

- Slug: garage-ai
- Year: 2026 · Role: Full Stack Developer · Status: Live
- Categories: SaaS, Voice AI
- Stack: React, Vite, Tailwind CSS, SMS automation
- Results: 95% missed-call detection; <3s sms response time; 47% lead conversion
- Links: [Case study](https://hamzakahloon.online/projects/garage-ai) · [Live site](https://www.garageai.co.uk) · [Source code](https://github.com/Hamza-Kahloon786/Garages_misscall_detect)

A lead-recovery SaaS for UK garages: every missed call triggers one polite SMS that guides the customer to a short form and delivers a ready-to-book lead.

Garages lose work every time the phone rings out. GarageAI catches each missed call, texts the caller instantly and walks them through a simple form.
The garage receives a clean lead — vehicle, registration and the job needed — without changing its phone number.

What was built:
- Real-time missed-call detection with an instant SMS follow-up, working with the garage's existing number.
- Guided lead form capturing vehicle, registration and service required.
- ROI calculator, transparent £49/month pricing and an admin dashboard.
- GDPR-compliant lead handling, built for the UK market.

### Intellex CRM — One AI inbox for every social account

- Slug: intellex-crm
- Year: 2025 · Role: Frontend & Product Engineer · Status: Live
- Categories: SaaS, AI & LLM
- Stack: React, WebSockets, AI auto-response, Lead scoring
- Links: [Case study](https://hamzakahloon.online/projects/intellex-crm) · [Live site](https://intellex-orcin.vercel.app) · [Source code](https://github.com/Hamza-Kahloon786/intellex)

A multi-account social-media CRM that unifies Facebook, Instagram and WhatsApp conversations, scores leads automatically and replies with AI.

Intellex pulls every conversation from a company's social accounts into one inbox, so no message — and no lead — slips through.
Each conversation gets a lead score, and AI auto-responses keep customers engaged around the clock.

What was built:
- Unified inbox across Facebook, Instagram and WhatsApp accounts.
- Automatic lead scoring on every conversation.
- AI auto-responses with real-time notifications over WebSockets.
- Overview dashboard: connected accounts, conversations, unread messages and qualified leads.

### AI Visibility Intelligence API — A 3-agent pipeline for AI-search visibility

- Slug: ai-visibility-api
- Year: 2026 · Role: Backend & AI Engineer · Status: Open source
- Categories: AI & LLM
- Stack: Python, Flask, OpenAI, Docker, pytest
- Links: [Case study](https://hamzakahloon.online/projects/ai-visibility-api) · [Source code](https://github.com/Hamza-Kahloon786/AI-Visibility-Intelligence-API)

A Flask REST API that profiles a business, runs a 3-agent AI pipeline to discover high-value AI-search queries, scores each opportunity and recommends content to close the gaps.

As more people search through AI assistants, businesses need to know which questions they should be the answer to. This API finds them.
Three agents work in sequence — discovering queries in the business's competitive space, scoring them for opportunity and generating content recommendations.

What was built:
- 3-agent AI pipeline: query discovery → opportunity scoring → content recommendations.
- Transparent opportunity-score formula, with real data and a heuristic fallback.
- Dockerised with Gunicorn; database migrations run automatically on boot.
- Documented API reference, prompt-engineering notes and a test suite.

### NexpreneurAI — Launch a business — no tech or English skills needed

- Slug: nexpreneur-ai
- Year: 2026 · Role: Full Stack Developer · Status: Live
- Categories: AI & LLM, SaaS
- Stack: React, Vite, Tailwind CSS, Node.js, Express, MongoDB
- Results: 10 interface languages
- Links: [Case study](https://hamzakahloon.online/projects/nexpreneur-ai) · [Live site](https://nexpreneurai.com) · [Source code](https://github.com/Hamza-Kahloon786/NexpreneurAI)

An AI platform that helps first-time founders launch and grow: describe an idea in your own language and get a business plan, product descriptions and more.

NexpreneurAI is built for people who have a business idea but not the technical or English skills to execute it.
Users pick their language, describe their idea, and the platform generates the documents they need to get started — then tracks their progress.

What was built:
- Pick from 10 interface languages on first visit — including Urdu, Arabic, Hindi, Bengali, Pashto, Somali, Turkish, Polish and Romanian.
- Describe a skill or idea in a few words and get a personalised business plan — product ideas, pricing suggestions, marketing strategy and next steps.
- AI tools that generate product listings, social-media content and visuals.
- Learning Hub with guides on selling and growing, plus FAQs and support.
- Progress dashboard for completed AI tasks and activity over time.
- Email / password and Google OAuth sign-in with protected routes.

### Virtual AI Companion — A 3D chatbot you can talk to

- Slug: virtual-ai-companion
- Year: 2024 · Role: Full Stack & AI Engineer · Status: Open source
- Categories: Voice AI, AI & LLM
- Stack: React, Three.js, FastAPI, NLP, Voice AI
- Links: [Case study](https://hamzakahloon.online/projects/virtual-ai-companion) · [Source code](https://github.com/Hamza-Kahloon786/r3f-Virtual-GF)

An interactive 3D AI companion with a voice interface — real-time speech-to-text and text-to-speech, contextual NLP responses and animated WebGL characters.

A 3D character rendered in the browser that listens, understands and talks back — combining WebGL, voice processing and NLP in one experience.

What was built:
- Interactive 3D companion rendered with WebGL / Three.js (React Three Fiber).
- Real-time voice processing with speech-to-text and text-to-speech.
- NLP models for contextual understanding and personalised responses.
- Responsive front end with video-call style interaction and real-time animations.

### Nozama.ai — AI-powered e-commerce marketplace

- Slug: nozama-ai
- Year: 2025 · Role: Full Stack Developer · Status: Case study
- Categories: AI & LLM, Machine Learning
- Stack: React, Node.js, MongoDB, Machine Learning
- Results: 96% recommendation accuracy
- Links: [Case study](https://hamzakahloon.online/projects/nozama-ai) · [Demo video](https://www.linkedin.com/posts/hamza-kahloon-12a14125a_fullstackdevelopment-aimarketplace-react-ugcPost-7369043559905136640-dOR7/)

A marketplace with ML-driven product recommendations, a dynamic pricing engine and a 24/7 AI support chatbot.

Nozama.ai brings AI to every step of shopping — what customers see, what they pay and how they get help.

What was built:
- ML-driven product recommendations with 96% accuracy.
- Dynamic pricing engine.
- 24/7 AI chatbot support for customers.

### RiskSim Enterprise — Monte Carlo risk simulation platform

- Slug: risksim-enterprise
- Year: 2025 · Role: Full Stack Developer · Status: Open source
- Categories: SaaS, Machine Learning
- Stack: React, TypeScript, Tailwind CSS, FastAPI, MongoDB, NumPy
- Results: 10,000 iterations per simulation
- Links: [Case study](https://hamzakahloon.online/projects/risksim-enterprise) · [Source code](https://github.com/Hamza-Kahloon786/risk_simulation)

Enterprise risk management with 10,000-iteration Monte Carlo simulations, an interactive scenario canvas and financial-impact analysis.

RiskSim helps organisations put a number on risk: model a scenario, run thousands of simulations and see the likely financial impact — and the ROI of defending against it.

What was built:
- 10,000-iteration Monte Carlo risk simulations powered by NumPy / SciPy.
- Interactive canvas for building risk scenarios.
- Financial impact, revenue-loss and security-ROI calculations.
- Multi-location risk monitoring, incident tracking and defence-effectiveness metrics.

### AI Call Center — Production call-center SaaS with AI voice

- Slug: ai-call-center
- Year: 2025 · Role: Full Stack Developer · Status: Open source
- Categories: Voice AI, SaaS
- Stack: React, Python, PostgreSQL, Voice AI
- Links: [Case study](https://hamzakahloon.online/projects/ai-call-center) · [Source code](https://github.com/Hamza-Kahloon786/Call_Center)

A scalable, production-ready call-center SaaS with a modern UI, a robust backend and AI-powered voice integration — built for real users, not just an MVP.

Designed from day one as a production system: a modern interface for teams, a backend built to scale, and AI voice built into the call flow.

What was built:
- AI-powered voice integration inside the call flow.
- JavaScript front end with Python services and a PostgreSQL data layer.
- Architected for scale and real users rather than a throwaway MVP.

### ProjectPilotHub — Software-studio site with an AI assistant

- Slug: project-pilot-hub
- Year: 2026 · Role: Full Stack Developer · Status: Live
- Categories: Web, AI & LLM
- Stack: Next.js, React, Node.js, Express
- Links: [Case study](https://hamzakahloon.online/projects/project-pilot-hub) · [Live site](https://project-pilot-hub.vercel.app) · [Source code](https://github.com/Hamza-Kahloon786/ProjectPilotHub)

Marketing site and API for a software studio offering FYP solutions, SaaS development, AI automation and ML/NLP services — built SEO-first with Next.js.

A fast, SEO-focused marketing site that presents the studio's services and work, with an AI assistant ready to answer visitors' questions.

What was built:
- Next.js App Router front end with a full SEO setup.
- Express API that emails contact-form submissions via Nodemailer.
- “Ask ProjectPilotHub AI” assistant available on every page.
- Services, portfolio, hire-a-developer and blog sections.

### Usman Laser Eye Clinic — Patient registration, tokens & appointments

- Slug: laser-eye-clinic
- Year: 2026 · Role: Full Stack Developer · Status: Open source
- Categories: Healthcare, Web
- Stack: React, Vite, Tailwind CSS, Node.js, Express, MongoDB
- Links: [Case study](https://hamzakahloon.online/projects/laser-eye-clinic) · [Source code](https://github.com/Hamza-Kahloon786/Laser-Eye-Clinic)

A MERN system for a real eye clinic: reception registers patients without duplicates and issues daily queue tokens; doctors work the queue and open full records.

Built around how the clinic actually works — a receptionist at the front desk and doctors moving through a daily queue.

What was built:
- Permanent, atomically generated MR numbers per patient (e.g. MR-000001).
- Search by name, phone or MR number to reuse existing records — no duplicates.
- Daily token queue with per-patient status flow for doctors.
- JWT authentication with Receptionist and Doctor roles; appointment management and stats.

### Actuarial Job Board — Scraped listings with a full CRUD dashboard

- Slug: actuarial-job-board
- Year: 2025 · Role: Full Stack Developer · Status: Open source
- Categories: Web
- Stack: Python, Flask, PostgreSQL, React, Selenium
- Links: [Case study](https://hamzakahloon.online/projects/actuarial-job-board) · [Source code](https://github.com/Hamza-Kahloon786/Scraping)

A full-stack job board that scrapes actuarial postings from Actuary List with Selenium and serves them through a Flask API and React UI with filtering and sorting.

Built as a hiring assessment for Bitbash — scraping, API and UI in one clean full-stack project.

What was built:
- Selenium scraper for automated job-data collection.
- Flask REST API with SQLAlchemy on PostgreSQL / MySQL.
- React UI to view, add, edit and delete jobs with filtering and sorting.

### Anxiety Detection in Urdu — NLP for mental-health signals in Urdu text

- Slug: anxiety-detection-urdu
- Year: 2022 · Role: ML Engineer · Final-year research · Status: Case study
- Categories: Machine Learning, AI & LLM
- Stack: Python, Streamlit, Flask, NLP, scikit-learn
- Results: 88% classification accuracy
- Links: [Case study](https://hamzakahloon.online/projects/anxiety-detection-urdu)

An NLP model trained on a custom Urdu dataset to detect anxiety in text, served through a Flask API with a Streamlit interface — 88% classification accuracy.

Most mental-health NLP work targets English. This project built a custom Urdu dataset and a classifier that detects anxiety in Urdu text in real time.

What was built:
- Custom Urdu dataset for mental-health analysis.
- Flask API for real-time text classification.
- Streamlit interface for interactive exploration.
- 88% anxiety-classification accuracy.

### Bird Sound Recognition — Species classification from audio

- Slug: bird-sound-recognition
- Year: 2025 · Role: AI/ML Engineer · Decoders Digital · Status: Case study
- Categories: Machine Learning
- Stack: Python, Machine Learning, Audio processing
- Links: [Case study](https://hamzakahloon.online/projects/bird-sound-recognition)

An AI/ML system that identifies bird species from their calls using audio signal processing and machine learning.

Raw field recordings are processed into features a model can learn from, then classified into species automatically.

What was built:
- Audio signal processing to turn raw recordings into model-ready features.
- Machine-learning classifier for automated species identification.

### Grain Hub — Brand website for a grain supplier

- Slug: grain-hub
- Year: 2025 · Role: Frontend Developer · Status: Live
- Categories: Web
- Stack: React
- Links: [Case study](https://hamzakahloon.online/projects/grain-hub) · [Live site](https://grain-hubs.vercel.app) · [Source code](https://github.com/Hamza-Kahloon786/GrainHubs)

A business website for a supplier of premium wheat, rice and flour — products, services, team and contact, with a hero carousel and a consent modal.

A clean, image-led brand site that presents the supplier's products and promise to customers.

What was built:
- Hero carousel with product storytelling.
- Products, services, team and contact sections.
- “Natural promise” consent modal on first visit.

## Skills

- **AI & LLM:** OpenAI, LangChain, LangGraph, Hugging Face, RAG, NLP, OCR, Gradio, Streamlit
- **Voice & Real-time:** Twilio, ElevenLabs, Vapi, Whisper, Deepgram, WebRTC, WebSockets
- **Frontend:** React, Next.js, Angular, TypeScript, Redux, Tailwind CSS, Three.js, Bootstrap, HTML5, CSS
- **Backend:** Node.js, Express, FastAPI, Django, Flask, Spring Boot, Laravel, REST APIs
- **Machine Learning:** scikit-learn, TensorFlow, PyTorch, OpenCV, Pandas, NumPy, Matplotlib, Power BI
- **Data & Vectors:** PostgreSQL, MongoDB, MySQL, Supabase, SQLite, Redis, Pinecone, Chroma, FAISS, Qdrant, Weaviate
- **Automation & DevOps:** n8n, Zapier, Docker, Git, GitHub, Vercel, Railway, AWS, Hostinger, pytest
- **Integrations:** Stripe, QuickBooks, Google Calendar, Google Maps, SendGrid, Mailgun, AWS SES, Cloudinary
- **Languages:** Python, JavaScript, TypeScript, Java, C++, PHP, SQL

## Certifications

- Web Development with AI/ML — Pangea Global Enterprise (PGE) (2025)
- Full Stack Development — Nexus AI Lahore
- Web Development — ITU University · Arfa Tower
- Web Designer & Developer — PNY Training (2024)
- UI/UX Design — Great Learning Academy (2024)
- Python Programming — SoloLearn (2023)
- AI/ML Fundamentals — SoloLearn (2023)
- C++ Programming — Great Learning Academy (2023)
- Advanced JavaScript — Coursera (2022)
- Database Management — MongoDB University (2022)

## Career timeline

- **2021 · First lines, first clients:** Started freelancing — building Java applications, including a complete hospital management system.
- **2022 · BS Software Engineering:** Graduated from The University of Lahore (CGPA 3.3). Final-year research: anxiety detection in Urdu text with NLP — 88% accuracy.
- **2022 · Pangea Global Enterprise:** Joined PGE as a Full Stack Developer (AI/ML) — enterprise multi-tenant apps, FastAPI microservices and ML integrations.
- **2023 · Specialising in AI:** Built AIPAS — procurement automation with OCR, NLP and RAG, reaching 92% accuracy on automated decisions.
- **2024 · Voice, 3D & agents:** Shipped voice-enabled AI experiences, NLP chatbots and LangChain / LangGraph agent workflows for production clients.
- **2025 · Decoders Digital:** Joined Decoders Digital. Launched LeadFront (formerly Vendira.ai) AI voice agents and STORM AI — a CRM serving 1,000+ daily active users.
- **2026 · AI for healthcare:** Led the multi-tenant, HIPAA-focused rebuild of Doctor AI Portal with AI patient intake and automated SOAP notes.

## Frequently asked questions

### Who is Hamza Kahloon?

Hamza Kahloon (full name Muhammad Hamza Tanveer Kahloon) is a Full Stack Developer & AI/ML Engineer based in Lahore, Pakistan. He has 4+ years of experience building AI-powered products — AI voice agents, RAG systems and multi-tenant SaaS platforms — and currently works at Decoders Digital.

### What does Hamza specialise in?

AI engineering and full-stack product development: LLM integrations with OpenAI, RAG pipelines, LangChain and LangGraph agents, real-time voice agents with Twilio, ElevenLabs, Deepgram and Whisper, and multi-tenant SaaS built with React, Next.js, Node.js, FastAPI and Django.

### What are Hamza's most notable projects?

Highlights include LeadFront, an AI front office of voice and SMS agents for service businesses; Doctor AI Portal, a multi-tenant healthcare SaaS with AI patient intake; STORM AI, a CRM serving 1,000+ daily active users; AIPAS (ProcureHub), procurement automation with 92% decision accuracy; and EZOS, a proposal and commission platform for sales teams.

### Which technologies does Hamza work with?

Python, JavaScript and TypeScript; React, Next.js and Angular on the front end; Node.js, Express, FastAPI, Django and Flask on the back end; OpenAI, LangChain, LangGraph and Hugging Face for AI; PostgreSQL, MongoDB, Supabase, Redis and vector databases such as Pinecone, Chroma and FAISS; and Docker, Vercel and AWS for deployment.

### Is Hamza available for hire?

Yes — open to full-time roles & freelance projects. He works with teams worldwide from Lahore, Pakistan and usually replies within 24 hours.

### How can I contact Hamza?

Email hamzaakahloon903@gmail.com, message on WhatsApp at +92 309 1453950, or connect on LinkedIn. The contact form on this site also opens a ready-to-send email or WhatsApp message.

### Where did Hamza study?

BS Software Engineering at The University of Lahore (2022, CGPA 3.3 / 4.0). His final-year research detected anxiety in Urdu text with NLP at 88% accuracy.