Skip to content
Muhammad Siraj

Founding AI Engineer at Eleven SoftwaresIslamabad, Pakistan

Muhammad Siraj, AI engineerI build AI agents that are measured before they’re trusted.

Agentic systems with LangGraph, RAG and MCP, built so the model interprets and tested code acts. 10+ client AI projects shipped, from architecture to production.

View workGet in touch
Portrait of Muhammad Siraj

Muhammad Siraj, Islamabad

  • LangGraph
  • LangChain
  • Model Context Protocol
  • RAG
  • CrewAI
  • FastAPI
  • PostgreSQL
  • ChromaDB
  • LangSmith
  • Pydantic
  • AWS Bedrock
  • Docker

Selected work

3 projects

  • The hvacX demo page, “Try the front desk”, with the MD Cold Services chat open beside a website enquiry form.

    AI front desk for HVAC

    hvacX

    A chat and email front desk that books HVAC visits with zero double bookings and answers questions with 99% correctness on its evaluation set.

    • 99% answer correctness
    • 0.846 MRR on a 51-question eval
    • Zero double bookings

    Stack: LangGraph, MCP, RAG, FastAPI, PostgreSQL, ChromaDB, OpenAI, Pydantic, APScheduler, pytest

  • The StormLead overview dashboard. “2 homeowners contacted after storms in 1 city. 2 inspections booked.” A severe-weather card for Dallas, TX, and a funnel from outreach sent to booked.

    Storm-triggered lead engine

    StormLead

    Four autonomous LangGraph pipelines that turn storms across 10 metros into outreach, replies and booked inspections, with a person approving every email before it sends.

    • 4 autonomous pipelines
    • 10 metros monitored
    • Tested on 400 seeded customers

    Stack: LangGraph, MCP, Composio, OpenAI, PostgreSQL, React, FastAPI, LangSmith

  • Title card for PrepLens showing a spoken question and answer exchange.

    AI interview preparation platform

    PrepLens

    A real-time voice interviewer, AI résumé analysis and instant-scored mock exams that cut manual evaluation by 70%.

    • 70% less manual evaluation

    Stack: React, FastAPI, PostgreSQL, Redis, OpenAI, Gemini, Whisper

About

I’m an AI engineer in Islamabad, and I build agents that businesses can hand real work to. As the founding AI engineer at Eleven Softwares I have shipped 10+ client AI projects end to end, from the first architecture sketch to production.

My working rule is simple: the language model decides what a request means, and tested code decides what happens next. So the systems I build are measured before they are trusted. Retrieval is evaluated, risky steps wait for a person’s approval, and guardrails stand between model output and anything that matters.

I study computer science at COMSATS University Islamabad on an HEC Top Merit Scholarship, where I also served as general secretary of the campus Google Developer Group.

Client AI projects shipped end to end
10+
Answer correctness on the hvacX evaluation
99%
Faster responses after query optimisation (60s to under 2s)
30×
Less manual work from meeting-lifecycle agents
70%
How I build agentsModel output is dashed. Code is solid.
  1. A request arrivesChat, email, a storm alert
  2. The model interpretsTyped, validated output
  3. Tested code actsBookings, emails, writes
  4. Confirmed after the writeNothing claimed early

Experience

  1. to

    Founding AI Engineer

    Eleven Softwares, Remote

    • Shipped 10+ client AI projects end to end, from architecture to production deployment.
    • Built agentic workflows with LangGraph, LangChain and CrewAI that automate the full meeting lifecycle, cutting manual work by 70%.
    • Integrated MCP servers that connect agents to ClickUp, Slack, Google Drive and Notion.
    • Optimized database queries, bringing response latency from 60s to under 2s (30× faster) through a user-centric, demo-driven process.

  2. to

    Full Stack Engineer

    Das Soft Ltd, Islamabad, on site (internship)

    • Built the D-Unit IoT platform with React, Express.js and Firebase for real-time (<200ms) power control, cutting monitoring by 70%.
    • Co-designed an optimized database schema, improving data flow and integration efficiency by 40%.

  3. to

    General Secretary

    Google Developer Groups on Campus, COMSATS University Islamabad, Hybrid (volunteer)

    • Co-organized a web development bootcamp with 1,000+ registrations and led technical sessions.
    • Led the DataCamp Scholarship Program, distributing 200 free licenses to students.
    • Certified instructor for the Web Bootcamp with LLM Workflows (May 2025), teaching MERN and Gemini integrations.

Skills

LLMs and agents
  • LangGraph
  • LangChain
  • CrewAI
  • MCP
  • OpenAI API
  • Tool calling
  • Pydantic
  • Prompt engineering
RAG and evaluation
  • ChromaDB
  • Embeddings
  • Vector search
  • LangSmith
  • Hit@k
  • MRR
  • LLM-as-judge
ML and voice AI
  • Hugging Face
  • scikit-learn
  • pandas
  • NumPy
  • AssemblyAI
  • Deepgram
  • ElevenLabs
Backend and databases
  • FastAPI
  • Node.js
  • Express.js
  • Celery
  • PostgreSQL
  • Redis
  • Firestore
Cloud and DevOps
  • AWS (EC2, S3, Lambda, Bedrock)
  • Docker
  • Kubernetes
  • GitHub Actions
  • Railway
  • Vercel
Frontend
  • React
  • Tailwind CSS
Languages
  • Python
  • JavaScript
  • SQL
  • Java
  • C++

Education

to

COMSATS University Islamabad

Bachelor of Science in Computer Science, Islamabad, Pakistan

  • CGPA 3.5 / 4.0
  • HEC Top Merit Scholarship for the full four-year program

Certifications

  • Nov 2026

    AI Engineer Production Track: Deploy LLMs and Agents at ScaleUdemy (Ed Donner)

  • Oct 2026

    1 to 1 Million WorkshopAWS Student Builder Group, COMSATS University Islamabad

  • Jul 2026

    AI Engineer Agentic Track: The Complete Agent and MCP CourseUdemy (Ed Donner)

  • Jul 2026

    AI Engineer Core Track: LLM Engineering, RAG, QLoRA, AgentsUdemy (Ed Donner)

Contact

Building something with agents that has to work every time? I’d like to hear about it.

msirajkazim@gmail.com
Based in
Islamabad, Pakistan
Résumé
Download PDF