Lahore, Pakistan · GMT+5 Open to new projects

AI systems that solve real business problems.

I’m Muhammad Siddiq. I build RAG systems, AI agents, and WhatsApp & voice automation, from the first conversation about what’s going wrong to a product running in production.

See the case studies Tell me your problem

What I build

01RAG & knowledge assistants
02AI agents & automation
03WhatsApp & voice agents
04LLM evaluation
05Backends & SaaS

Evaluation

Proof, not a demo.

My client assistants ship with an evaluation suite. These numbers come straight from the scorecards: the first two from the Edge Manpower Assistant, 460 from StackMinds, still in progress.

01Latency

4.8s

p95 reply time on WhatsApp, against a 5 s target

02Cost

$0.00017

Cost per answered question

03Coverage

460

Curated test cases gating each release

01 — Work

Problems I’ve solved

Every project opens as a short case study: the problem, my plan, and the outcome.

Clinic WhatsApp Agent

Problem: A doctor's WhatsApp was flooded with bookings, questions and emergencies. Result: Parents book in Urdu around the clock, and emergencies reach the doctor at once.

A WhatsApp assistant for a paediatric clinic. It books and manages appointments, answers questions about the clinic, and hands urgent or sensitive conversations to the doctor, who replies through the same bridge.

Edge Manpower Assistant

Problem: A recruitment chatbot that guesses a salary or visa rule does real harm. Result: Answers only from verified facts: 100% on correctness and safety in evaluation.

A WhatsApp assistant for Edge Manpower, an overseas recruitment company. It answers candidates from the company's knowledge base, refuses to invent what it doesn't know, and ships with an evaluation suite that measures every stage.

StackMinds Recruiting Agent

Problem: Hiring, sales and follow-up chats all landed on one WhatsApp number. Result: One assistant sorts them, screens candidates and emails HR a finished summary.

A WhatsApp-native assistant for StackMinds that does three jobs in one conversation: recruits candidates, answers company, pricing and support questions from a knowledge base, and handles follow-ups. This project is still in progress.

OpenGPT

Problem: Teams want ChatGPT on their own server, with their own documents. Result: A self-hosted assistant with tools, document chat and memory that lasts.

A ChatGPT-style assistant you can run yourself: streaming answers, conversation threads that persist, chat with your own documents, and an agent that picks the right tool for the question.

GraceWise

Problem: Valuable guidance was buried in PDFs nobody had time to search. Result: A live platform where parents simply ask, and answers come from the sources.

A children's spiritual growth platform for parents and educators. An AI guide answers questions from a curated library of source documents, and a curriculum system delivers structured lessons by age and learning stage.

ElevateXCrew — website & CMS

Problem: Every blog post or job opening needed a developer. Result: A fast, SEO-ready site the team now updates themselves.

The corporate website and content platform for ElevateXCrew, taken from business requirements to production: a server-rendered public site, a REST API and a dedicated admin portal.

RaketH Clone — AI voice SaaS

Problem: A voice AI engine can't sell itself: no accounts, billing or admin. Result: A complete SaaS around it: credits, voice library and a back office.

The complete SaaS layer for an AI voice engine: it connects to a separate GPU voice backend and wraps voice cloning, multilingual speech and translated voice in everything needed to run them as a business.

LMS Agentic

Problem: School software buries every task under menus and forms. Result: An assistant that acts for you, limited to what your role allows.

A full learning management system for administrators, teachers and students, with a LangGraph assistant that carries out tasks using tools matched to the role of whoever is asking.

YouTube RAG Chat

Problem: Finding one answer in a two-hour video means scrubbing and guessing. Result: Ask the video directly and get answers grounded in its transcript.

A Chrome extension that turns any YouTube video into something you can question. It reads the transcript, indexes it, and answers from what was actually said.

OpenWA — document support

Problem: WhatsApp bots could see a document arrive, but couldn't open it. Result: Files now reach the webhook, ready for AI and RAG pipelines.

OpenWA is an open-source, self-hosted HTTP gateway for WhatsApp. I extended it so documents sent by users actually arrive in the webhook and can be downloaded on demand.

02 — Approach

Problem first. Model second.

Most AI projects fail because they start with a framework. Mine start with what is costing you time or money.

Understand the problem

I map who is stuck, where the manual work is, and what “fixed” looks like. If AI isn’t the right answer, I’ll say so.
Plan the system

A clear architecture before any code: the data flow, the right model, where retrieval, tools and people fit, and how it fails safely.
Ship and measure

A production backend, a real deployment, and evaluation results, so you can see it working instead of trusting a demo.

03 — Experience

Where I’ve worked

Feb 2026 – Present Current role
Generative AI Engineer

ElevateXCrew · Remote
- Design and deploy agentic AI pipelines, RAG systems, AI chatbots and n8n automation workflows for client projects, cutting manual operations.
- Built voice-to-voice AI agents with STT/TTS, an AI appointment-booking chatbot, and a client project monitoring system as internal products.
Jan 2025 – Oct 2025
Generative AI Trainer

Tecrix · On-site
- Designed and delivered hands-on curricula on Generative AI, prompt engineering and LLM orchestration frameworks.
- Mentored students and developers through real RAG pipelines, API integrations and vector databases.
- Supervised capstone projects on autonomous agents and workflow automation, from first concepts to deployed systems.
Dec 2024 – Jan 2025
AI/ML Engineer

DevelopersHub Corporation · Remote
- Designed and implemented LLM-based RAG systems and chatbots for workflow automation and customer knowledge retrieval.
- Built AI-driven automation including social media auto-posting and n8n workflows.

04 — Skills

Tools I work with

01
AI & LLMs
- LLMs
- RAG pipelines
- Agentic AI
- Voice-to-voice agents
- MCP
- NLP
- AI automation
02
Frameworks
- LangChain
- LangGraph
- LangSmith
- OpenAI API
- Hugging Face
- n8n
- Twilio
03
Backend
- Python
- FastAPI
- Flask
- Pydantic
- SQLAlchemy
- Node.js
- Prisma
04
Databases & vectors
- PostgreSQL
- MySQL
- Redis
- Supabase
- Pinecone
- FAISS
- Chroma
05
Frontend
- React
- Next.js
- TypeScript
06
Machine learning
- Scikit-learn
- TensorFlow / Keras
- NumPy
- Pandas
- Jupyter
07
Deployment
- DigitalOcean
- Railway
- Vercel
- Netlify
- Hostinger
- cPanel
- Git & GitHub
08
AI coding tools
- Claude Code
- Cursor
- GitHub Copilot
- OpenCode
- Kilo Code
- Gemini CLI
- Qwen CLI

Certifications

01Oct 2025 GenAI Engineer Tecrix
02Aug 2025 Backend Engineer Akhuwat FIRST, Faisalabad

Education

01 BS Computer Science Virtual University of Pakistan
02 Intermediate Akhuwat College Kasur

05 — Contact

Got a problem worth solving?

Tell me what’s slow, manual or broken. I’ll reply with how I’d approach it: a plan, not a sales pitch.

muhammadsiddiq.code@gmail.com

GitHub LinkedIn Download CV

AI systems that solve real business problems.

What I build

Proof, not a demo.

Problems I’ve solved

More case studies

Problem first. Model second.

Understand the problem

Plan the system

Ship and measure

Where I’ve worked

Generative AI Engineer

Generative AI Trainer

AI/ML Engineer

Tools I work with

AI & LLMs

Frameworks

Backend

Databases & vectors

Frontend

Machine learning

Deployment

AI coding tools

Certifications

Education

Got a problem worth solving?