class MuhammadFarhan:
role = "Senior Python, AI and Cloud Engineer"
location = "Lahore, Pakistan | Open to global remote"
edge = "I architect systems, not just write scripts"
approach = "Agentic AI pipelines + battle-tested cloud infra + fast delivery"
stack = {
"ai_agents" : ["LangGraph", "CrewAI", "Claude API", "Claude Code", "MCP", "OpenAI"],
"backend" : ["Django", "FastAPI", "Celery", "Redis", "Async IO"],
"data" : ["PostGIS", "HuggingFace", "Pinecone", "ETL Pipelines"],
"cloud" : ["ECS Fargate", "Lambda", "Aurora PG", "CDK (TS)", "OIDC CI/CD"],
"frontend" : ["React.js", "Next.js", "TypeScript", "Tailwind CSS"],
}
superpower = "Hand me any stack, any domain. I will ship."I don't just prompt AI. I build systems where AI does the work.
Most engineers treat AI as an autocomplete. I treat it as a runtime: agents that plan, delegate, use tools, and close the loop without hand-holding.
What this looks like in production:
- LangGraph: stateful agent graphs with conditional routing, memory layers, and multi-step tool-use across long-horizon tasks
- CrewAI: role-based agent crews for research, structured extraction, and parallel reasoning workflows
- Claude API + MCP: context-aware agents with custom tool registries, long-context reasoning, and structured JSON outputs
- Claude Code (terminal): AI-native dev environment for shipping features, writing tests, debugging, and refactoring at full speed directly from the terminal
- Cursor + Codex: sprint-length tasks compressed into hours; entire modules scaffolded in minutes
- RAG Pipelines: Pinecone and pgvector, from raw document ingestion to sub-100ms semantic search in production
- Prompt Engineering: few-shot chains, chain-of-thought orchestration, output parsers with 85%+ verified production accuracy
Work that used to take months now runs in hours.
Not marketing. Check the commit history.
🔒 HIPAA-compliant deployments · Blue/Green with auto-rollback · No SSH, no long-lived credentials · Encryption at rest and in transit
|
Mine Rehabilitation Tracking and 3D GIS Visualization Platform Multi-tenant SaaS for mining operations worldwide: tracks rehabilitation progress, visualizes mine data in 3D, and gives operators a single source of truth across sites. I led the entire product from zero: client discovery sessions, Figma design direction, React frontend, Django backend, GIS data engineering, and HuggingFace model deployment for lightweight inference. Built the ETL pipeline that ingests complex mine GIS datasets into the platform and architected the multi-tenant data layer so each mining company's data stays isolated and secure. The one project where I owned every layer of the stack.
🔗 mapmymine.com |
||
|
Concrete Fleet Activity Intelligence Used at the Wilshire Grand: the largest concrete pour in history: 2,000+ truckloads over 20 hours. I built the activity prediction engine: GPT-4 few-shot prompting on GPS telemetry, drum rotation sensors, and speed data to classify each truck's state (loading, transit, pouring, washout) at 85%+ accuracy. The live map dashboard feeds directly into automated, billing-ready job records for clients including Concrete Supply Co. and Firth Concrete. 
🔗 truckast.com |
AI Stakeholder Intelligence (Netherlands) Policy and advocacy teams were spending months manually mapping stakeholder networks. I integrated OpenAI APIs for stakeholder identification and influence scoring, then built a RAG pipeline on top. What took months now takes a single query. FastAPI wrapper deployed on AWS EC2. 
🔗 dialog.nl |
AI Creator Outreach Platform Django REST Framework powering TikTok Shop and Creator Auth API integrations at scale. Owned the full AWS deployment (EC2, S3, RDS) and rebuilt the Celery task pipeline to push throughput up 60% through async parallelism. Thousands of creator campaigns running without intervention. 
🔗 hivehq.ai |
|
GIS Parcel Visualization Platform Real-time geospatial parcel analytics serving property and land management teams. PostGIS spatial queries, GeoDjango, and async Celery pipelines for report generation. Cut processing time 40%. High-availability AWS EC2 and RDS deployment behind NGINX.
🔗 mapport.com |
UK Full-Fibre ISP: GIS Network Tooling Built a QGIS Plugin in Python and PyQt5 for the network planning team: snapping tools, polygon merging, and multi-threaded processing for large fibre route datasets. Mentored junior developers. Promoted to project lead within 7 months. 
🔗 countybroadband.co.uk |
Affiliate Marketing Automation Django and PostgreSQL backend with OpenAI-powered campaign intelligence and Discord bots handling real-time creator engagement. Celery handles weekly partner API sync automatically: no manual intervention needed.
🔗 bemomentiq.com |
|
GIS Agricultural Mapping Platform AWS S3 upload and processing pipelines via Celery, REST APIs for GIS data visualization and precision agriplot mapping. EC2 deployment with Gunicorn and cron jobs for scheduled data refreshes.
🔗 agriplot.earth |
||
