I build LLM-powered agent systems, RAG pipelines, and the scalable services that keep them honest in production.
Currently an AI & Cloud Engineer at AI Marketplace GmbH and an M.Sc. Computer Science candidate at Paderborn University, working on multi-agent orchestration, explainable AI, and knowledge graphs.
Experience
Nov 2024 to Present
AI & Cloud Engineer
AI Marketplace GmbH · Working student, 20 hrs/week · Paderborn
Agent Engine: scalable microservices
Delivered a 3x improvement in service response times by building a Redis-backed scaling system, without any loss in output quality.
Designed a 3-mode orchestrator agent, Plan, Ask, and Accept All, giving users fine-grained control between fully autonomous runs and step-by-step review.
Added support for custom LangGraph workflows, so providers can extend the platform beyond what ships out of the box.
Multi-Agent 3D Generation Framework
Led the architecture of a 13+ agent framework that automatically creates, edits, and removes 3D structures, a new take on multi-agent systems.
Microsoft Copilot RAG System, Client: prostep ivip
Led end-to-end design and deployment of a Copilot assistant with role-based access, handling 3,000+ documents and serving users directly inside MS Teams.
Built a sandboxed Python runner that executes user code inside isolated Docker containers, closing off remote code execution risks.
Oct 2023 to Nov 2024
Student Research Assistant
DICE Research Group · Paderborn University
Benchmarked 6 open-source LLMs (Mistral, Mixtral, Llama 2/3, Zephyr) and reached 90%+ accuracy on material-science tasks after fine-tuning, matching top proprietary models.
Mapped hundreds of research papers and material-science references into the Springer knowledge graph, improving RAG workflow efficiency by 25%.
Set up and enriched a SPARQL endpoint over the graph using Wikipedia substance and property data, lifting data organization quality by 30%.
Nov 2020 to Dec 2022
Python Instructor
Cue Learn & Self-employed · India · 80+ students, 4 countries
Taught Python and small-game development (pygame, tkinter) to 80+ students across the US, UK, Canada, and Switzerland, with a 95% satisfaction rate.
Earned Star Teacher recognition within 3 months of joining Cue Learn, the fastest in the cohort.
Publication
ESWC 2024 · Posters & Demos · Springer LNCS
CLASS MATE: Cross-Lingual Semantic Search for Material Science Driven by Knowledge Graphs
First project of its kind for cross-lingual search in material science. Built a knowledge graph with multilingual labels for chemical substances, developed a similarity-based entity recognition method, and shipped a demo that retrieves information across multiple languages.
A. Perevalov, J. Chinchghare, M. Krishna, S. Sharma, A. N. Lal, A. Deshwal, A. Both, A.-C. Ngonga Ngomo
Pitching at Makathon #XCHANGE4INDUSTRY · Paderborn · Dec 2025
Makathon #XCHANGE4INDUSTRY
Fraunhofer IEM · Team of 5 ·1st place, €3,000
Solved two real-world problems for Werkzeugbau Berger GmbH: a high-temperature sensor module for injection moulds that enables predictive maintenance, and Digital Vault, a shared document platform with versioning used by Berger, their suppliers, and their clients.
Vibathon · Google Developer Group · Paderborn · Dec 2025
Vibathon
Google Developer Group ·1st place
Built an AI agent that calls doctors for the user, books appointments, and syncs them to the calendar, end-to-end and in any language.
Projects
Fact-Checking Engine over Wikipedia KG
RDF · SPARQL · Python
Validation engine leveraging the Wikipedia knowledge graph, achieving 92% accuracy on claim verification.
Explaining GNNs on Heterogeneous Graphs
PyTorch Geometric · Captum
Explainability system for graph neural networks reaching 88% explanation accuracy in predicted links.
Requirement Doc Generation with Generative AI
FastAPI · Noctua HPC
Led backend for a 5-person team; cut project runtime by 70% via code optimization and fine-tuning on Paderborn's supercomputer.
ASL Learning with Computer Vision
OpenCV · MediaPipe
Gesture-recognition app teaching American Sign Language at 95% detection accuracy; led a team of three.
Education
Apr 2023 to Present
M.Sc. Computer Science
Paderborn University · Paderborn, Germany
Machine Learning · Explainable AI · Knowledge Graphs · Large Language Models · Deep Learning · Unsupervised Learning (R). Master's thesis in progress.
Aug 2017 to Jun 2021
B.Tech Computer Science & Engineering
KCC Institute of Technology & Management (AKTU) · India
