Amro Moustafa

about me über mich

Robotics, Automotive & AI Backend Engineer (Python).

M.Sc. in Mechatronics & Robotics Engineering (Leibniz Universität Hannover) — specialized in multi-body dynamics, robotics kinematics, control systems, and sensor fusion.
5+ years cross-domain engineering experience spanning robotics & cobots (Fraunhofer IPA), dynamic modeling & reinforcement learning (IAV GmbH), and autonomous systems validation & enterprise AI platforms (EDAG Group).
Human-Robot Collaboration (HRC) & Safety: Engineered collaborative robot simulation suites, cycle-time models, and dynamic risk-zone analyzers conforming to ISO/TS 15066 at Fraunhofer IPA.
Physical AI & Control Optimization: Applied DDPG & PPO Reinforcement Learning for closed-loop controller tuning (overshoot reduced 15% → 4%) and sub-second autonomous decision pipelines.
Enterprise AI Platform Architect: Designed and deployed production RAG & document intelligence systems (Python/FastAPI, Celery, PostgreSQL/pgvector, Azure OpenAI & local Qwen3 embeddings) at EDAG.
Model-Based & HIL Automation: Built automated CI/CD and HIL/SiL validation toolchains (MATLAB/Simulink, TPT, dSPACE, CARLA, Vector tools) cutting test cycle times by 25%.
Languages & Certifications: Arabic (Native) · German (C1+ Professional) · English (C1+ Professional) · ISTQB® CTFL 4.0 certified.

email

amro.mo.moustafa@gmail.com

phone telefon

+49 1590 1355023

location standort

Berlin, Germany

Years in
Robotics & Automotive

25%

Faster CI/CD
Test Cycles

Faster Defect
Triage (LLM+ML)

OEM Vehicle
Platforms

Agentic AI
Pipelines

companies i have worked with

experience berufserfahrung

EDAG Group — embedded across Tier 1 OEM programmes CURRENT

Systems Integration & Validation Engineer / Enterprise AI Architect — Autonomous Systems, HIL & AI Platforms

05/2023 – Present

Berlin, Germany

Enterprise AI document-intelligence platform
built a Python/FastAPI + Celery backend (four task queues) with PostgreSQL/pgvector, Redis and S3 that ingests SAP C4C sales attachments for OCR, Azure OpenAI summarisation/classification and RAG chat, surfaced in a Next.js/React frontend.
Full-stack debugging & LLM prompt engineering
traced a silent cross-stack regression (React → FastAPI → Celery) that dropped generated tables; shipped a disciplined three-layer fix and reverted premature workarounds; resolved an LLM context-inversion hallucination via document-classification tags and added a 262-line test suite.
Legacy Office & export pipeline overhaul
implemented .doc/.xls/.ppt processors (OLE2 → OOXML → HTML fallbacks, no LibreOffice) and registered them in the processor registry; replaced LibreOffice with WeasyPrint and python-docx for brand-compliant DOCX/PDF export plus a live PDF-preview React component.
Prompt architecture & scalable ops
DRY'd 12 shared rule blocks via a <<PLACEHOLDER>> _inject() mechanism across 20+ system prompts (zero runtime overhead); operated a GPU-bounded sequential Celery worker in a distributed India/Germany team on Bitbucket, Jenkins CI, Kubernetes/Helm and Keycloak.
CARLA autonomous simulation & scenario validation
built an automated CARLA 0.9.14 simulation toolchain (Python, Docker, AWS EC2 GPU) for closed-loop ADAS scenario execution, sensor injection, KPI evaluation, and automated Jira Xray REST API traceability.
End-to-end software integration ownership
served as technical anchor for ADAS validation programmes across Tier 1 OEM customers; responsible for the full pipeline from requirements decomposition through branch coordination, integration test execution, defect triage, and release evidence delivery; held direct accountability for integration quality at programme milestone reviews.
CI/CD pipeline design & quality gates
architected scalable GitLab CI / Docker pipelines (Python + Bash), enforcing smoke tests, regression coverage, and automated quality-gate criteria across multiple vehicle software platforms; 25 % reduction in integration cycle time; reused without rework across three major VW software releases.
Software release engineering & SDLC coordination
coordinated software content delivery across development, test, and project teams; managed Git branching strategies aligned to release milestones and SDLC phase exits; owned test case readiness, defect lifecycle status tracking, and quality gate compliance; reported integration health and release readiness to OEM programme stakeholders weekly.
Requirements & test traceability
maintained bidirectional traceability from safety goals through system requirements to test cases and validation evidence in DOORS and Jira Xray; automated coverage-matrix generation via Jira Xray REST API; zero untraced requirements at every programme milestone review.
Integration dashboards & observability
built CI/CD-integrated dashboards (GitLab, Python, Jira Xray) aggregating real-time integration health, quality KPIs, and test campaign status across multiple software builds; automated alerting on integration failures and quality-gate violations.
Defect triage & root-cause analysis
led structured triage of integration and regression failures with feature teams; isolated defects across ECU software, communication stacks, and test infrastructure; produced high-fidelity root-cause reports and drove corrective actions to closure; translated findings into concise summaries for programme management.
Cross-functional integration leadership & SDLC governance
chaired integration planning meetings, defect lifecycle review calls, and release decision gates across ADAS software, hardware, network, and test infrastructure teams; managed stakeholder escalations and drove conflict resolution for integration blockers; defined test plan entry / exit criteria and enforced release QA standards; mentored junior engineers in Git branching, CI/CD practices, and structured test processes.
HIL / SiL simulation & CARLA-based validation toolchain
designed closed-loop test sequences on HIL benches (dSPACE, Speedgoat) for ADAS ECU validation; built a complete CARLA 0.9.14 simulation toolchain (Python, Docker, AWS EC2 GPU) automating scenario execution, KPI evaluation, and Jira Xray REST API traceability updates; worked daily with CAN, LIN, and Ethernet at integration layer using Vector tools and Wireshark; managed ECU flash workflows and software update validation sequences.

IAV GmbH

Function Developer / Application Engineer

09/2021 – 03/2023

Gifhorn, Germany

Automated data integration tooling
built Python + Bash pipelines for .a2l / .hex ECU calibration file ingestion via INCA; automated feature extraction, sensor deviation correction, and structured database population (MySQL, InfluxDB) for time-series vehicle measurement data; zero manual steps after initial deployment.
Model-based development & validation
developed and validated Simulink / Stateflow models for vehicle air management systems; applied control theory, system identification, and statistical verification against CAN logged measurement data; maintained UML diagrams and DOORS style documentation for cross-team audit traceability; ensured model fidelity at SiL level before handover to HIL integration.
AI-driven controller optimisation
applied DDPG Reinforcement Learning to autonomously optimise controller parameters; achieved overshoot reduction from 15 % → 4 % over 50,000 training steps with formal convergence validation and documented performance evidence; implemented in Python / TensorFlow with structured test reports reviewed by engineering leadership.
Requirements engineering & measurement data analysis
decomposed vehicle system requirements into testable specifications; maintained traceability from system requirement to validation evidence; built pandas / asammdf data processing pipelines for multi-channel time-series analysis; produced structured reports for cross-functional review sessions including function owners and release management.

Fraunhofer IPA

Research Engineer — Simulation & Automation

07/2018 – 04/2019

Stuttgart, Germany

Simulation & automation tooling
developed simulation and automation tools for industrial manufacturing scenarios: cycle-time modelling, safety-zone analysis, and robot / sensor configuration evaluation; worked with multi-disciplinary teams across research and production-floor contexts.
Safety zone & risk assessment modelling
developed simulation models for safety zone definition in human-robot collaboration scenarios, applying risk assessment methodology (ISO 10218 / EN ISO 13849) to establish safe operational boundaries — directly analogous to ODD and MRC analysis used in autonomous driving functional safety.
Robot & sensor evaluation tool
engineered a Java application for systematic robot and sensor selection (cost, time, automation potential); designed SolidWorks 3D models of mechanical assemblies and produced manufacturing drawings and documentation to optimise production workflows and quality assurance.
Cross-functional research collaboration
integrated design, evaluation, and implementation phases within interdisciplinary teams spanning research and industrial automation; maintained project documentation on Confluence and managed version-controlled codebases with Git throughout the research lifecycle.

BMW Group / Mercedes-Benz Egypt

Automotive Diagnostics Trainee

2013 / 2014

Egypt

ECU diagnostics & safety verification
performed functional safety-system verification and ECU diagnostics on production-ready vehicles using OEM diagnostic platforms (CAN / LIN); interpreted DTC fault codes, live data, and test logs to distinguish hardware faults from software / configuration defects; supported final inspection and test-drive quality gates.
Automated test plan execution & sensor calibration
assisted technicians in executing systematic automated test plans using BMW diagnostic platforms; validated sensor calibration and verified the functionality of safety-critical and comfort systems across integrated electrical / electronic E / E architectures.
ECU software flashing & DTC fault-finding
applied OEM diagnostic tools over CAN / LIN to read and clear DTCs, perform control-unit tests, and execute guided troubleshooting; observed and supported ECU software update and flashing procedures, reinforcing understanding of software load validation and post-flash verification on commercial vehicles.

projects & systems projekte & systeme

Enterprise AI / Platform Engineering

Enterprise AI Document-Intelligence Platform

Robotics, Automotive & AI Backend Engineer (Python)

Python 3.11 FastAPI Celery PostgreSQL / pgvector Redis / S3 Azure OpenAI Qwen3 Embeddings RAG Next.js / React Kubernetes / Helm

Autonomous Systems / HIL & CARLA

Autonomous Scenario & HIL Validation Toolchain

Built a complete ADAS/autonomous vehicle scenario validation toolchain from scratch: CARLA 0.9.14 simulation engine, Python orchestration, Docker containerisation, AWS EC2 GPU fleet management, GitLab CI pipeline, and Jira Xray REST API integration. Reduced integration cycle time by 25% across three major OEM vehicle platforms with zero rework.

CARLA 0.9.14 Python HIL / dSPACE Docker AWS EC2 GPU GitLab CI Jira Xray

Human-Robot Collaboration & Safety Zones

Robotics / Human-Robot Collaboration

Cobot Simulation & Dynamic Safety Zones (ISO/TS 15066)

Engineered simulation suites at Fraunhofer IPA for collaborative robot (cobot) workcells: modeled dynamic cycle times, reachable workspace envelopes, and risk zones under ISO/TS 15066 & ISO 10218. Developed a Java/C++ application benchmarking industrial manipulators and sensor suites, paired with SolidWorks 3D kinematics assemblies.

Robotics ISO/TS 15066 Java / C++ Kinematics SolidWorks 3D Cobots

Control Theory / Reinforcement Learning

Autonomous Controller Tuning via Deep RL (IAV GmbH)

Applied DDPG Reinforcement Learning to autonomously optimize closed-loop dynamic controller parameters at IAV GmbH. Achieved overshoot reduction from 15% → 4% over 50,000 training steps. Modeled multi-body dynamic systems in MATLAB/Simulink and verified against INCA calibration data (.a2l/.hex).

DDPG / RL MATLAB / Simulink Python INCA (.a2l/.hex) CAN Analysis

Automotive Systems / HIL & Simulation

CARLA Autonomous Systems & HIL Validation Toolchain

Architected a high-fidelity autonomous driving scenario validation toolchain combining CARLA 0.9.14 (Python, Docker, AWS EC2 GPU) and HIL test benches (dSPACE, Speedgoat). Features automated scenario execution, closed-loop sensor injection, real-time KPI evaluations, and bidirectional Jira Xray REST API traceability.

CARLA 0.9.14 HIL / SiL (dSPACE) Python Docker Vector CANoe ISO 26262

Physical AI / Autonomous Decision Systems

Autonomous RL Decision & Optimization Pipeline

Architected an autonomous decision-making pipeline utilizing Proximal Policy Optimization (PyTorch PPO) in custom Gymnasium environments. Incorporates sub-second streaming feature extraction, state-space vectorization, live risk circuit breakers, and fault-tolerant inference loops for high-frequency control.

PyTorch PPO Gymnasium Real-Time Telemetry Circuit Breakers Low Latency

Multi-Agent Systems / Knowledge Graph

Dynamic Hierarchical Multi-Agent Graph Engine

Built a deterministic multi-tier agent orchestration framework integrating Neo4j knowledge graphs, Model Context Protocol (MCP), and guarded execution lanes. Features automated multi-agent role decomposition, persistent memory retrieval, and strict token-budget optimization for autonomous engineering tasks.

Neo4j Multi-Agent Systems MCP Protocol Python Graph Memory

hi , i am AmroMoustafa. hi , ich binAmroMoustafa.

about me über mich

Robotics, Automotive & AI Backend Engineer (Python).

email

phone telefon

location standort

experience berufserfahrung

EDAG Group — embedded across Tier 1 OEM programmes CURRENT

IAV GmbH

Fraunhofer IPA

BMW Group / Mercedes-Benz Egypt

projects & systems projekte & systeme

Enterprise AI Document-Intelligence PlatformEnterprise-KI-Dokumenten-Intelligenz-Plattform

Autonomous Scenario & HIL Validation ToolchainAutonome Szenario- & HIL-Validierungstoolchain

Cobot Simulation & Dynamic Safety Zones (ISO/TS 15066)Cobot-Simulation & Dynamische Sicherheitszonen (ISO/TS 15066)

Autonomous Controller Tuning via Deep RL (IAV GmbH)Autonome Regleroptimierung via Deep RL (IAV GmbH)

CARLA Autonomous Systems & HIL Validation ToolchainCARLA Autonome-Systeme & HIL-Validierungs-Toolchain

Autonomous RL Decision & Optimization PipelineAutonome RL-Entscheidungs- & Optimierungs-Pipeline

Dynamic Hierarchical Multi-Agent Graph EngineDynamische hierarchische Multi-Agenten-Graph-Engine

key achievements wichtige erfolge

End-to-End Integration Toolchain Built from Scratch End-to-End-Integrations-Toolchain von Grund auf aufgebaut

AI-Accelerated Defect Triage in Production KI-beschleunigte Defekt-Triage in der Produktion

Zero Untraced Requirements — Every Milestone Null ungetrackte Anforderungen — jeder Meilenstein

Fleet-Scale Integration Health Dashboards Flottenweite Integrations-Gesundheits-Dashboards

SOTIF Scenario Coverage Architecture SOTIF-Szenario-Coverage-Architektur

Agentic AI Document-Intelligence Platform — Architected & Shipped Agentische KI-Dokumenten-Intelligenz-Plattform — architektiert & ausgeliefert

skills & core proficiencies fähigkeiten & kernkompetenzen

92%

90%

95%

95%

92%

95%

92%

90%

education ausbildung

10/2020 – 03/2023

M.Sc Mechatronics & Robotics Engineering

Leibniz Universität Hannover

Hannover, Germany — Grade: 2.2

10/2019 – 10/2020

M.Eng Mechatronics

Hochschule Merseburg

Merseburg, Germany

09/2009 – 07/2014

B.Sc Mechanical Power Engineering

Mansoura University

Mansoura, Egypt

certifications zertifikate

languages sprachen

contact me kontakt

email

phone telefon

location standort

hi , i am
Amro
Moustafa. hi , ich bin
Amro
Moustafa.

Enterprise AI Document-Intelligence Platform

Autonomous Scenario & HIL Validation Toolchain

Cobot Simulation & Dynamic Safety Zones (ISO/TS 15066)

Autonomous Controller Tuning via Deep RL (IAV GmbH)

CARLA Autonomous Systems & HIL Validation Toolchain

Autonomous RL Decision & Optimization Pipeline

Dynamic Hierarchical Multi-Agent Graph Engine

End-to-End Integration Toolchain Built from Scratch

AI-Accelerated Defect Triage in Production

Zero Untraced Requirements — Every Milestone

Fleet-Scale Integration Health Dashboards

Agentic AI Document-Intelligence Platform — Architected & Shipped