Applied AI Engineer · Available for work
I build agents that decide, retrieval your team can trust, and the Python backends holding it all upright — delivered as working software, not a proof of concept.
LangChain, LangGraph and hand-written ReAct loops with real tool routing — SQL toolkits, live web gateways, deterministic calculators. The agent picks the path; the system keeps it honest.
Outcome — work that consumed hours of manual effort now runs unattended.
RAG pipelines tuned against real corpora, not toy datasets. Vector stores, semantic search and chunking strategies chosen for the documents you actually have.
Outcome — answers grounded in your data, with the receipts to prove it.
FastAPI services, PostgreSQL, ETL pipelines, OAuth2 and JWT, Docker, background workers. The unglamorous layer that decides whether any of it ships.
Outcome — a system that survives real traffic, not just a demo.
Selected work
Each one holds the screen as you scroll — the outcome it delivered, how it was built, the stack behind it, and the repository.
PropTech · Autonomous agent
An enterprise PropTech pipeline paired with a tool-using AI analyst. It harvests unstructured global property listings through headless Playwright automation, normalises them in a Pandas warehouse and persists via SQLAlchemy. On top sits a zero-framework ReAct loop bound to Groq's Llama 3.3 70B, routing execution between SQL toolkits, live web gateways and deterministic financial calculators.
E-commerce · Multi-tool agent
An autonomous AI analyst for e-commerce. Aisle ingests store data into a consolidated warehouse, then places a multi-tool agent on top of it. A native tool-calling loop on Groq/Llama 3 wields twelve distinct tools to forecast demand, recommend pricing and segment customers through RFM analysis, streaming responses token by token into a premium Next.js dashboard.
Security · Autonomous extraction
An autonomous agent that takes instructions in natural language and extracts structured, schema-validated data from any website. A two-stage fetch keeps token costs down by deciding what actually needs a full render, extraction schemas are generated at runtime, and every resolved IP is validated against loopback, private and cloud-metadata ranges before a request is made.
Aerospace · Air-gapped systems
An enterprise-grade, fully offline client–server architecture. It bypasses the cloud entirely by establishing an asynchronous HTTP polling network across a local subnet, coordinating exam delivery and collection between machines that never touch the internet. A Pandas ETL engine ingests messy CSV question banks and normalises them into a clean SQLite store behind a PySide6 desktop interface.
Data engineering · Automation
An end-to-end data engineering pipeline. It extracts e-commerce pricing data via Python Requests, normalises it and enforces strict typing through Pandas so malformed rows never reach storage, then loads everything into a relational SQLite database through SQLAlchemy. Secure FastAPI endpoints serve the result to a Chart.js dashboard for real-time comparison.
Data engineering · Warehousing
Extracts live job market data from isolated sources that share no common format — REST APIs, BeautifulSoup web scrapers and static CSV exports — then cleans and reconciles it under strict Pandas logic. Everything is unified into a single PostgreSQL data warehouse and served through FastAPI to a responsive Tailwind analytics dashboard.
Backend · Data-as-a-Service
An enterprise FastAPI backend for secure, high-concurrency asynchronous data processing. It implements strict OAuth2 and JWT authentication, Alembic migrations and a domain-driven design structure that keeps the codebase navigable as it grows. BackgroundTasks paired with Pandas absorb large CSV workloads without ever tying up a request thread.
Capabilities
Grouped by the layer of the system each tool actually lives in.
Layer 01
Layer 02
Layer 03
Experience
1 yr+ · Current
Upwork · Global clients
A 100% job success score held across the engagement. Delivered LLM agent orchestration systems, automated extraction pipelines and production APIs — replacing manual reporting workflows with deterministic Python logic.
4 months
Inovaqo
Optimised backend infrastructure in Python and FastAPI alongside the engineering team, cleaned datasets, wrote the SQL layer, and gained hands-on commercial experience deploying scalable applications.
Client reviews
Verified feedback from Upwork clients and direct enterprise engagements.
"Ibrahim delivered an exceptional piece of software architecture for our facility. We required a fully offline examination network with zero internet connectivity due to strict security protocols. He engineered a flawless Python backend — highly recommended."
AKAMSC PAC K.Director of Training, NASTP
"Ibrahim built exactly what we needed and delivered it on time. He took the time to understand our requirements before writing any code, which meant the agent worked properly from day one."
UCUpwork ClientAI Agent Development · LangChain / RAG
"I have a great time working with Muhammad Ibrahim — he is quick to the solution, fast, easily accessible, and the work is delivered flawlessly."
UCUpwork ClientData Crawling & Web Scraping
"Work was delivered exactly as we needed it. Passionate and eager to learn about the project scope, requirements and domain."
UCUpwork ClientSenior Backend & AI Engineering
Available for new work
Freelance projects and full-time roles alike — RAG systems, autonomous agents, or the ETL and API layer holding it all together.