● Interactive CV · Full-stack AI Developer & Technical Product Manager · Hamburg, Germany

Farivar Zarvandeh — I turn businessintent intointelligent systemsthat ship.

I'm Farivar Zarvandeh. 13+ years from telecom platforms serving millions of users to agentic AI for modern enterprises — working where business strategy becomes engineering execution.

See my projects
A small floating white robot with a big screen face and paddle hands — the site's companion (static version)
Experience
13+ years
Products I built, 2024–2025
8, full-stack
Focus
Agentic AI · Automation · Scalable Data Products
Portrait of Farivar Zarvandeh

01 — Resume

Farivar Zarvandeh

Full-stack AI Developer & Technical Product Manager

Full-stack AI Developer and Technical Product Manager with 13+ years in telecom and enterprise software. I turn business goals into shipped AI products — from 0-to-1 MVP to scale — and stay hands-on with LLMs, agentic workflows and cloud architecture.

Experience
13+ years
Built 2024–2025
8 products, full-stack
Location
Hamburg, Germany

Professional experience

  1. Lead AI Product Manager & Solutions Architect

    Broszio & Co. · Hamburg, Germany ·

    • Delivered the company's first AI agent layer, automating customer-service workflows with LLMs.
    • Built a speech-to-search platform (LangChain, vector databases) for natural-language queries on internal data.
    • Led an Amazon sales-forecasting engine that improved inventory-planning accuracy.

  2. Head of AI Product Labs (Tech Lead)

    MTN Irancell · Tehran, Iran ·

    • Founded and led AI Solution Labs for Iran's largest data operator.
    • Cut product delivery cycles by 30% with AI-assisted development (Copilot strategies).
    • Shipped GenAI tools for document processing, content generation and operational analytics.
    • Defined cloud-native architecture on Azure and Vertex AI.

Core competencies

AI & Technical Strategy

  • GenAI & LLM application lifecycle
  • RAG architectures
  • Agentic workflows
  • Solution architecture
  • Cloud strategy (GCP / Azure)
  • MLOps
  • System integration

Product Management

  • Product strategy & roadmap
  • Agile / Scrum leadership
  • MVP to scale
  • Stakeholder management
  • User-centric design
  • KPI definition
  • E-commerce strategy

Technical Stack

  • Python (Pandas, NumPy, FastAPI)
  • LangChain
  • OpenAI API
  • Vector databases
  • SQL
  • PostgreSQL
  • Docker
  • CI/CD
  • Git
  • BigQuery
  • Xentral ERP
  • Shopify

02 — My projects · 2024–2025

Eight products I built myself, end to end.

My own portfolio from 2024–2025. I conceived, designed and developed every product here full-stack — product strategy, architecture, backend, frontend, AI and deployment. Each one shows its real mechanism, its real interface where one exists, and an honest maturity label.

  1. Orders from Amazon, Shopify, POS, the online shop and eBay stream into the real Prodcommerce dashboard, guided by an LLM for listings, forecasts and pricing, then flow out to the warehouse, shipping and a DATEV export
    Core competencySelf-hosted multichannel ERP for merchants selling everywhere
    Beta

    ProdCommerce (Prodsystem)

    Designed & built full-stack by me

    ProdCommerce brings orders from Shopify, Amazon, Shopware, WooCommerce, eBay and POS into one order book. It then runs the whole back office on top of it, from warehouse dispatch to legally compliant German e-invoicing and DATEV export.

    • Unified order book across 6 sales channels plus POS
    • Scan-driven pick, pack and ship with DHL and Sendcloud labels
    • ZUGFeRD/XRechnung e-invoicing with DATEV export
    • Approval-gated returns, refunds and invoice corrections

    Source-agnostic back office. Every channel converges on one order upsert, so adding a connector requires no change to dispatch, invoicing or reports.

    • Python
    • Flask 3
    • SQLAlchemy 2
    • APScheduler
    • Gunicorn (gthread)
    • Pydantic
    Open case study

03 — My skills

What I can do, made tangible.

The seven disciplines I work in. Each one wakes up when you interact with it and shows how I apply that skill, with evidence from my own projects.

  1. 01Signals become reasoning

    Agentic AI & LLM Systems

    GenAI application lifecycle, RAG, agentic workflows and multi-provider LLM integration — with guardrails and human approval where it matters.

    • GenAI & LLM lifecycle
    • RAG architectures
    • Agentic workflows
    • LangChain / LangGraph
    • OpenAI · Anthropic · Gemini APIs
    • Vector databases
    • Voice AI (ElevenLabs)

    evidence → AI Agent Layer & Speech-to-Search at Broszio & Co.; LangGraph Magic Importer (Prodcommerce); multi-provider script & image pipeline (StoryBrook); hook-governed coding agents (ProdEngine).

    ◆ The network you just woke up resolves into the retrieval pipeline this site's own digital representative uses: query → retrieve → ground → answer.

  2. 02Services find their boundaries

    Solution Architecture

    Cloud-native, integration-heavy system design: clear service boundaries, idempotent flows and message-driven integration with external platforms.

    • Solution architecture
    • System integration
    • Cloud strategy (GCP / Azure)
    • Event & outbox patterns
    • Multi-tenant design
    • API design

    evidence → Cloud-native architecture on Azure & Vertex AI (MTN Irancell); six-channel order integration with one idempotent upsert (Prodcommerce); outbox/EDI design (ProdBusiness).

    ◆ Hover a service: its real dependencies light up, everything else steps back. Boundaries are a design decision, not an accident.

  3. 03Chaos settles into schema

    Data Engineering & Databases

    Relational modelling, data products and analytics — from PostgreSQL invariants and row-level security to BigQuery and forecasting.

    • SQL
    • PostgreSQL
    • Row-level security
    • BigQuery
    • Pandas · NumPy
    • Forecasting (LightGBM, statistical)
    • Data products

    evidence → 480-table PostgreSQL model with 474 RLS policies (ProdBusiness); Amazon sales forecasting engine (Broszio & Co.); FBA ensemble forecasting (Prodcommerce).

    ◆ Scattered records snap into tables; the query beam joins exactly the rows it needs.

  4. 04An interface is a stack of decisions

    Full-stack Product Engineering

    Shipping complete products across the stack — Python, TypeScript and PHP backends, React/Next.js frontends, VS Code extensions and PWAs.

    • Python · FastAPI · Flask
    • TypeScript · React · Next.js
    • Node · Express
    • PHP
    • VS Code extension API
    • Three.js
    • PWA

    evidence → Flask ERP (Prodcommerce), PHP e-invoicing SaaS (Prodinvoice), Express + React studio (StoryBrook), Next.js PWA (JobXentral), VS Code extension (ProdEngine).

    ◆ The card separates into the layers that make it work — interface, state, API, data — then snaps back together.

  5. 05Commits become deployments

    Cloud, DevOps & MLOps

    Reproducible delivery: containers, CI/CD, validation in the pipeline and deployment automation.

    • Docker & Compose
    • CI/CD (GitHub Actions)
    • Git
    • MLOps
    • Azure · GCP · Vertex AI
    • Coolify / SSH deploys

    evidence → CI with real DB + KoSIT validator (Prodinvoice); veraPDF in CI (Prodcommerce); Docker/Coolify deploy panels (ProdEngine).

    ◆ Each stage only passes what it verified. Red stops the line — on purpose.

  6. 06Manual steps become a system

    Process Automation

    Turning repetitive operational work — advertising, warehousing, data entry — into automated, observable pipelines.

    • Workflow automation
    • Scheduled jobs
    • ERP integration (Xentral)
    • Shopify
    • Document processing
    • Speech-to-search

    evidence → Advertising & warehousing pipelines (Broszio & Co.); ~30 scheduled jobs (Prodcommerce); AI document processing tools (MTN Irancell).

    ◆ Watch the hand-offs disappear: five manual steps collapse into one autonomous loop with a human checkpoint.

  7. 07Vague goals become a roadmap

    Product Leadership

    Product strategy, roadmapping and Agile leadership — from discovery to deployment, with KPIs that tie engineering to business value.

    • Product strategy & roadmap
    • Agile / Scrum
    • MVP to scale
    • Stakeholder management
    • KPI definition
    • E-commerce strategy

    evidence → Founded AI Solution Labs (MTN Irancell); 30% shorter delivery cycles; sprint-planned product repos (33-sprint plan in Letterman, 14 delivered sprints in ProdEngine).

    ◆ Scattered ideas get weighed, sequenced and committed: discovery → MVP → scale.

04 — How I work

Four rules I actually build by.

Not slogans — each one shows up in the projects above.

  1. 01

    Prove it, then scale it

    A working prototype beats a slide deck.

    EvidenceProdBusiness and JobXentral run as working prototypes.

  2. 02

    Rules live in the system

    If it matters, the code enforces it — not a checklist.

    EvidencePostgreSQL guards in ProdBusiness · fail-closed validation in Prodinvoice · task-gated edits in ProdEngine.

  3. 03

    AI proposes, humans approve

    Agents show their evidence; people make the call.

    EvidenceApproval gates in ProdBusiness, StoryBrook and Prodcommerce.

  4. 04

    Outcomes over architecture

    Every technical choice must pay off in speed, cost or revenue.

    Evidence“I don't just design systems; I ensure they deliver ROI.”

05 — My digital representative

Ask my CV itself.

A voice-first AI sidekick that knows Farivar's work and loves brainstorming yours. Tell it about your business and it pitches real AI ideas on the spot. Speak in English, Deutsch, فارسی, Français, Español, Türkçe or العربية; switch to text any time.

I'm Farivar's AI sidekick — I know his work and I'm here to brainstorm yours. I'm an AI, not Farivar himself. No audio plays until you start.

  • Creative

    Bring any business — a sandwich shop, a clinic, a warehouse — and get tailored AI and automation ideas, not a brochure.

  • Honest

    Facts about Farivar come from this site's own knowledge base; ideas are labelled as ideas. No made-up prices, clients or promises.

  • Next step

    Like an idea? Leave your details to set up a call with Farivar — or get a free, complete proposal within 24 hours.

06 — Contact me

Let's build the system that should exist.

Open to conversations about AI product leadership, solution architecture and agentic automation.

  • Based in Hamburg, Germany. Replies usually come from Farivar personally.

Your message is stored only to reply to you. No newsletter, no tracking.