AI systems that do real work.

I build production AI systems end to end: RAG assistants, MCP integrations, ElevenLabs voice agents and n8n automation pipelines, along with the FastAPI backends and Next.js frontends that put them in front of users.

My background is agency client work, turning scattered business tools into a single AI-powered workflow. I am looking for an AI Engineer role building LLM-powered products.

Timeline

(01)

Experience

Python Developer

Axioware Solutions

Sep 2025 - Present
Karachi, Pakistan

Designed and deployed AI-powered automation with n8n across 15+ business workflows, reducing manual operations by 70% across client accounts.
Built backend services with FastAPI, PostgreSQL and Redis, including JWT authentication, rate limiting and auto-scaling infrastructure on AWS.
Integrated GoHighLevel, Salesforce and the Meta Graph API with OpenAI and Claude into unified data pipelines, maintaining 99.5% processing accuracy across multiple enterprise accounts.
Shipped production ElevenLabs voice agents backed by FastAPI tool endpoints and an MCP server, deployed on AWS EC2 for live, real-time conversations.

n8n
FastAPI
PostgreSQL
Redis
AWS
OpenAI
Claude
ElevenLabs
MCP

(02)

Education

BS Software Engineering

Usman Institute of Technology

2022 - Feb 2026
CGPA 3.12 / 4.00

Final year project

UIT University website

Built on Next.js with a Sanity.io CMS so non-technical staff can create pages, edit copy and publish posts without a developer.

Read the case study Visit the site

(03)

Skills

AI & Automation: LLMs
RAG
LLM agents
MCP
n8n
OpenAI API
Claude API
ElevenLabs
Prompt engineering
Vector search (pgvector)
Backend: Python
FastAPI
Node.js
Express
REST APIs
JWT auth
Frontend: Next.js
React
TypeScript
Tailwind CSS
HTML/CSS
Cloud & Data: AWS (EC2, S3, CloudFront)
Linux
CI/CD
PostgreSQL
Redis
Supabase
Integrations: GoHighLevel
Salesforce
Meta Graph API
Asana
Fathom
Maqsam
Google Docs API

(04)

How I work

1
Start from the workflow, not the model
I map where the information actually lives and who needs it, then decide what the LLM should do. Most of the value is in the plumbing around it.
2
Ship the thin slice first
One working path end to end, in production, early. It surfaces the real problems while they are still cheap to fix.
3
Automate the boring parts
If a task is done by hand more than twice a week, it becomes a workflow. That is where most of the 70% reduction in manual operations came from.

Want the details? Ask me anything.

Get in touch See the work

Experience

Education

Skills

How I work

Start from the workflow, not the model

Ship the thin slice first

Automate the boring parts

Want the details? Ask me anything.