Munich, Germany

MuhammadAnas —builds AI systems

Full-stack & AI engineer. Production LLM agents with tool calling and RAG, the TypeScript and Python services around them, and the AWS infrastructure underneath.

Years shipping
2+
Saved / week
10h
Licensors priced
111

What I build

01 — Retrieval

Finding the right context

Documents become vectors. A question becomes a vector too. The neighbourhood that lights up is what the model actually gets to see — get this wrong and nothing downstream matters.

  • pgvector
  • Pinecone
  • LangChain
  • RAG

02 — Extraction

Messy in, schema out

HR emails and scanned receipts flow through extract, validate, store. What the model isn't sure about gets flagged for a human instead of quietly guessed at.

  • Python
  • FastAPI
  • Structured output
  • PostgreSQL

03 — Agents

Tools, not chat

A router reads the question and hands it to the specialist agent that can answer it. Twelve typed tools underneath. Agents propose changes; a person approves them, applied in one transaction, logged.

  • OpenAI Agents SDK
  • Responses API
  • Tool calling
  • NestJS

04 — Shipped

Running in production

Three internal apps at ProSiebenSat.1, a pricing engine across 111 licensors, and a 200-endpoint platform. All on AWS, provisioned in Terraform, released on merge.

  • ECS Fargate
  • Aurora
  • Terraform
  • GitLab CI/CD

Where I have shipped

Experience

  1. Full-Stack AI Engineer

    Redseven Entertainment GmbH(ProSiebenSat.1 Group)

    Munich, Germany · Mar 2026 — Present

    • Shipped three internal web applications in NestJS, Next.js and PostgreSQL that replaced Excel- and email-based workflows for TV production teams, saving them 10+ hours a week of manual data entry
    • Designed "Nellie", an AI assistant that lets rights and finance staff ask licensing cost questions in plain English — a fast routing model reads each question and hands it to the specialist agent that can answer it, built on the OpenAI Responses API
    • Wrote the tool-calling loop behind those agents, giving them 12 typed tools for cost breakdowns and missing licensing data, and stored conversations in PostgreSQL so users resume days later
    • Hardened the assistant for finance use: agents propose changes instead of writing them, and each approval is re-checked, applied in one database transaction, and logged. Role-based permissions block unscoped bulk edits
    • Developed LLM extraction pipelines turning HR emails and scanned receipts into schema-validated records, flagging fields the model is unsure about instead of guessing at them
    • Implemented the pricing engine reproducing legacy Excel cost rules across 111 licensors with live exchange rates, and integrated Microsoft Graph (Outlook, Entra ID SSO)
    • Deployed all three to AWS on ECS Fargate with Aurora Postgres and S3, provisioned in Terraform, with Docker builds and GitLab CI/CD releasing on merge

  2. AI Engineer

    Arcpeak

    Munich, Germany · Aug 2025 — Feb 2026

    • Built an AI business-analysis tool in Python on the OpenAI GPT APIs that reads a company's spend data and points out where enterprise clients are losing money
    • Created the conversational side with the OpenAI Agents SDK, giving the assistant tools it could call, guardrails on what it would answer, and Redis Streams for resumable sessions — users reached an answer 60% faster than with the old report-based flow
    • Replaced manual deploys by defining the AWS setup in Terraform (ECS Fargate, RDS, ElastiCache, ALB) and wiring GitHub Actions to build and release automatically, cutting deployment time by 80%
    • Added JWT and OAuth 2.0 login and Stripe subscription billing, enabling the company to onboard its first paying customers
    • Set up automated testing with Pytest, wired into the CI pipeline so tests run before anything ships
    • Delivered the React and TypeScript dashboard where clients read those insights and track their own metrics

  3. Backend Engineer (Freelance)

    Boardd(Enterprise Business Platform)

    Remote · May 2025 — Present

    • Built the backend for a business-management platform in Node.js and Express on MongoDB, growing it to over 200 REST endpoints covering projects, team collaboration, and client billing
    • Handled the money side with Stripe Connect so the platform could pay several parties at once, including Treasury accounts, virtual card issuing, recurring invoices, and onboarding for connected accounts
    • Made the app collaborative in real time over Socket.IO, with live task editing, Kanban boards, and drag-and-drop backed by optimistic locking, plus presence indicators showing who else is in a project
    • Secured it with JWT auth, rotating refresh tokens, Redis-backed sessions, and a role system with 40+ permissions
    • Connected messaging and storage services: Twilio SMS, SendGrid email, Firebase push, and file sync across AWS S3, Google Drive, Dropbox, and OneDrive

  4. Full-Stack Software Engineer

    WorkSpin

    Karachi, Pakistan · Jul 2023 — May 2024

    • Developed the backend for an event-discovery app in Node.js, using Socket.IO for live updates and adding OAuth/JWT login and Stripe checkout
    • Cut query latency by 70% and peak database load by 50% by reshaping the MongoDB schemas and adding Redis caching. Also delivered a gym-management system with AWS S3 media storage, live streaming, and Firebase alerts

Built on my own time

Projects

InsightQL

AI Database Assistant

A Next.js and NestJS tool that lets non-technical users query a database by typing a question in plain English, using LangChain's SQL agent over OpenAI GPT. Gets people an answer roughly 3x faster than writing the SQL themselves.

  • Next.js
  • NestJS
  • LangChain
  • OpenAI GPT

bugSage

AI Debugging Assistant

A FastAPI chatbot that pulls relevant docs and past issues out of a Pinecone vector database (RAG) before answering, so its fixes for Express.js bugs match the code you are actually running.

  • FastAPI
  • Pinecone
  • RAG
  • PyTorch

CLI Assistant

Agentic Terminal Tool

A Python assistant that runs entirely offline on a local model via Ollama, with an agentic loop that chains five tools together through function calling.

  • Python
  • Ollama
  • Function calling

The toolset

Skills

Languages
PythonTypeScriptJavaScriptSQLJava
AI & LLM
OpenAI APIOpenAI Agents SDKLangChainAI AgentsTool/Function CallingRAGVector Databases (pgvector, Pinecone)Prompt EngineeringPyTorch
Backend
Node.jsNestJSExpressFastAPIREST APIsGraphQLWebSockets (Socket.IO)PrismaTypeORM
Frontend
ReactNext.jsReduxTailwind CSSHTML/CSS
Databases
PostgreSQLMongoDBRedis
Cloud & DevOps
AWS (ECS Fargate, Aurora/RDS, S3, Secrets Manager)DockerTerraformKubernetesGitLab CI/CDGitHub ActionsGit
Practices
Agile/ScrumCode ReviewUnit Testing (Jest, Pytest)CI/CDMicroservices

Education

Studied

  • University of Passau

    MSc in Computer Science

    Passau, Germany · Oct 2024 — Present

  • National University of Computer and Emerging Sciences (FAST-NUCES)

    BS in Software Engineering

    Karachi, Pakistan · Aug 2020 — Jun 2024

Published

Research

Deep Learning for User Mobility Prediction in RIS-Assisted 6G THz Networks

IEEE

Benchmarked deep learning models for predicting user movement in next-generation (6G) mobile networks, to keep connections stable as users move.

Open to Full-Stack and AI roles

Let’s talk

Download CV (PDF)