Applied AI Engineer · Agentic Systems, Multimodal AI & Cloud Architecture
Egemen Gulpinar
Bielefeld, Germany
M.Sc. Intelligent Interactive Systems @ Bielefeld University
Where it all began ~2003
About Me
I am an Applied AI Engineer and Technical Lead with 5+ years of experience taking AI systems from idea to production. I lead AI, automation, research and cloud end-to-end across 25 live production services, with pipelines processing 50+ TB of data.
My focus is on agentic and multimodal AI: multi-agent platforms with MCP tool servers, agentic RAG and GraphRAG, video understanding across audio and visuals, and generative pipelines that turn days of expert work into minutes. Every system goes through evaluation, testing, benchmarking and monitoring before it reaches production, and is built with security, least-privilege access and GDPR (DSGVO) / EU AI Act requirements in mind, for on-premise or cloud deployment.
I also own cloud architecture and cost optimization across AWS, Azure and GCP. My philosophy is simple: embrace innovation, challenge the ordinary, and always aim to "be different."
Technical Expertise
Languages
Agentic AI
Core AI & ML
Data & Backend
Cloud
Methodologies & Tools
Certifications
- Certificate of Attendance, ESCAN 2026, Rome (attended with our published Frontiers paper)
- AZ-104T00: Microsoft Azure Administrator
- AI-102T00: Designing and Implementing Azure AI Solutions
- NVIDIA Inception Program - Deep Learning Institute
- Meta Back-End Developer Professional Certificate (Coursera)
- Inzva Applied AI Study Group
- Languages: Turkish (Native), English (Fluent, C1), German (B1, in progress)
- BAYKAR Technologies - AI Competition in Transport Certificate
Experience
5+ years building solutions across AI, backend systems, and research
AI Solutions Architect (Freelance)
LIVAD Technologies Shareholder
Los Angeles, CA (Remote)
Lead AI, automation, research and cloud end-to-end across 25 live production services, with pipelines processing 50+ TB of data and evaluation, testing, benchmarking and monitoring gates before every release.
- Video Understanding: Built a multimodal video-understanding system combining visual and speech analysis to enable text-to-search across thousands of videos, tracking emotion shifts and pinpointing key moments and laughter with precise timestamps, and feeding an agentic workflow that assembles ready-to-publish clips and articles.
- Multi-Agent Platform: Architected a multi-tenant AI agent platform (Claude Agent SDK, LangGraph) with MCP tool servers, an agent-native CLI and least-privilege security (scoped tokens, rate limiting, audit trails), plus an internal assistant built on agent skills and subagents that automates data, DevOps and support across Slack, Linear and Discord.
- LLM/SLM/VLM Fine-Tuning: Fine-tuned LLMs, SLMs and VLMs (Phi, Gemma, Qwen, GPT, LLaMA) on millions of data points for context recognition, and built VLM-based streamer face detection with real-vs-fake classification.
- Brand Safety: Created a GARM-aligned brand-safety and campaign-verification system combining STT, sentiment analysis and context translation, matching spoken campaign phrases word-by-word with timestamps at near-exact accuracy.
- Computer Vision: Built a multi-purpose asset and obstacle recognition and segmentation system using Segment Anything, feature extraction and custom algorithms.
- Cloud & Cost Optimization: Own cloud cost optimization across AWS, Azure and GCP, cutting GPU inference costs by 93% and continuously re-tuning workflows for efficiency; built CI/CD microservices and cloud tooling (GCP VM/container dashboard, data-lineage observatory).
- Awards: Led LIVAD Studio to win IAB MIXX Awards '23 Gold and Brandverse Awards '23 Silver for In-Game Ads.
Project Technical Lead
Bielefeld University (GusDaF Chatbot) with Türkisch-Deutsche Universität (TDU Istanbul)
Bielefeld, Germany (Hybrid)
- Leading an AI-driven educational system funded by BMFTR and DAAD and used by over 1,000 users, running open-source LLMs via GWDG on de.NBI Cloud for scalable, cost-efficient, low-latency real-time inference.
- RAG & OCR: Built a RAG pipeline that OCRs scanned documents into vector-DB-ready formats and automatically generates learning materials, questions and content from thousands of source data points.
- Learning Material Generation: Built a German learning-material generator covering all four skills (listening, reading, writing, speaking), with level-specific content, TestDaF exam preparation and near-natural multi-voice TTS.
Technical Lead
AI Agent for Mechanical Engineering
Bielefeld, Germany (Hybrid)
- Agentic RAG: Delivered an agentic RAG workflow for a large industrial enterprise client, with know-how transfer over hundreds of thousands of OCR-processed documents and tool-calling agents that analyze engineering schematics, built to GDPR (DSGVO) and EU AI Act requirements for on-premise or cloud deployment.
- Text-to-Drawing: Built a generative pipeline that produces production-ready technical drawings (EPLAN) from images, PDFs or plain text, successfully integrated into the client's engineering systems and cutting work that took engineers days down to minutes.
- Award: Winner of the Founders Foundation Pitch Day (Dec 2025).
Student Assistant - AI Research
Bielefeld University
Bielefeld, Germany
- Full-stack development of dialogue-based Explainable AI (XAI) system
- Built chunk-wise streaming support for LLMs to deliver low-latency, real-time outputs
- Enhanced dialogue system with real-time voice recognition (STT/TTS) and context-aware LLM functionalities
AI Engineer Intern
TUBITAK Space Technologies
Ankara, Turkey
- Researched Image Super-Resolution using Deep Convolutional Networks (SRCNN)
- Compared Input, Bicubic, and SRCNN results with PSNR and SSIM metrics
R&D Engineer Intern
BAYKAR Technologies
Istanbul, Turkey
- Satellite mapping, labelling and matching on World Map for BeiDou system
- Created special library for BeiDou RINEX data conversion, parsing and reading
- Discovered new relationship between BeiDou RINEX data and satellite movements
Education
M.Sc. Intelligent Interactive Systems
Bielefeld University, Germany
Sep 2024 - PresentErasmus Exchange - Computer Science
Bradford University, UK
Sep 2021 - May 2022GPA: 4.0/4.0
B.Eng. Computer Engineering
Mersin University, Turkey
Aug 2017 - Feb 2022GPA: 3.4/4.0
Projects
Selected work and open source contributions
GitHub Profile
All my open source projects, contributions, and experiments.
View All RepositoriesReal-Time Translation Keyboard
Native iOS keyboard that translates as you type across virtually all languages, in real time with near-zero latency. App Store release pending.
Language Learning Companion
Mobile-ready web app with daily vocabulary and phrase reminders and practice sessions for foreign-language learners. Release pending.
LIVAD Studio
In-game detection and advertising platform. Won IAB MIXX Awards '23 Gold and Brandverse Awards '23 Silver.
Explainable AI Dialogue System
Full-stack dialogue-based XAI system with real-time voice recognition and context-aware LLM interactions.
Image Super-Resolution
Deep Convolutional Network for image super-resolution. TUBITAK Space Technologies research.
BeiDou Satellite System
RINEX data parsing library and satellite position calculation for BeiDou navigation system.
Snap Camera-Kit WebSDK
Integration guide for Snap's AR Camera Kit in web applications. Includes Next.js deployment with Vercel.
Publications
Research papers and academic contributions
Emotional Modulation of Temporal Density in Natural Grasping: A Vision-Based Kinematic Study
Frontiers in Psychology — a peer-reviewed open-access journal published by Frontiers, covering research across all areas of psychology, including perception, cognition, and movement science.
doi.org/10.3389/fpsyg.2026.1823532Awards & Recognition
Founders Foundation Pitch Day Winner
AI-Based Digital Electrical Engineering Assistant - Technical Lead (Dec 2025)
IAB MIXX Awards '23 - Gold Award
In-Game Ads - Turkey & Europe Finalist
Brandverse Awards '23 - Silver Award
In-Game Ads Category
