I'm a job-ready AI Engineer specializing in Retrieval Augmented Generation (RAG), Large Language Models, and Agentic AI systems. Currently pursuing my Master's in Data Science at Berliner Hochschule für Technik (BHT) while working as an AI Engineer at Fraunhofer FOKUS.

With 3.5+ years of production-grade AI development and hands-on experience building enterprise RAG systems, I bring proven expertise in reducing hallucinations by 68%, improving retrieval precision, and deploying scalable AI solutions with Docker and CI/CD pipelines.

🎯 What makes me job-ready:

• ✅ Production experience with enterprise RAG systems at Fraunhofer FOKUS
• ✅ Microsoft Azure AI Certified (AI-900) - June 2025
• ✅ Proven track record with excellent employment references
• ✅ Strong German language skills (A2.1 level, actively learning)
• ✅ Full-stack AI development: from research to deployment

🎯 Enterprise RAG with Small-to-Big Retrieval & Hybrid Search

• Reduced hallucinations by 68% (38%→12%) using parent-child chunking with Docling HybridChunker
• Improved recall by 12% and precision by 8% with hybrid search (dense + BM25 sparse retrieval)
• Built evaluation framework with DeepEval: 0.82 answer relevancy, 0.88 faithfulness
• Integrated GitLab CI/CD with 7 quality gates, 88% pass rate, P95 latency <3.2s

🗣️ German Political Discourse Analysis with GottBERT

• Processed 193K speeches and 58 manifestos for semantic evolution analysis
• Developed two-stage neighbor extraction with temporal semantic visualization
• Deployed production tool: semantische-analysen.oeffentliche-it.de

Tech Stack: Docling, Qdrant, LlamaIndex, Docker, CI/CD, BERT, Transformers, Python

🎥 Watch Demo | Tech: GCP, Document AI, BigQuery, Vector Search, Vertex AI

• Built end-to-end RAG processing non-searchable PDF invoices with 95%+ field accuracy
• Implemented semantic search with 768-dimensional embeddings and 512-token chunking
• Deployed Vector Search index with top-3 cosine similarity retrieval

Tech: Docling, Qdrant, LlamaIndex, Docker, CI/CD

• Parent-child chunking architecture reducing hallucinations by 68%
• Hybrid search (85% semantic + 15% keyword) improving recall by 12%
• Production deployment with automated quality gates and regression tests

Tech: GottBERT, Transformers, PCA/UMAP, Plotly, Dash

• Processed 193K speeches with extended GottBERT vocabulary (+350 tokens)
• Multi-GPU inference with hierarchical caching and temporal visualization
• Deployed public tool with 143K data points

🏫 Berliner Hochschule für Technik (BHT) | Oct 2023 - Present
Master's in Data Science | Berlin, Germany
Specialization: Statistics, Machine Learning, Deep Learning, NLP, Text Mining

🏫 Maulana Abul Kalam Azad University of Technology | 2017 - 2021
Bachelor of Technology | India

🇩🇪 Deutsch Intensivkurs A2.1 | Jan 2026 - Feb 2026
Förde-vhs, Kiel | Course: Schritte Plus Neu 3

skills = {
    "AI_Engineering": ["RAG Systems", "LLM Fine-tuning", "Agentic AI", "Prompt Engineering"],
    "Machine_Learning": ["Supervised Learning", "Time Series", "NLP", "Ensemble Methods"],
    "Development": ["Clean Code", "OOP", "CI/CD", "Docker", "FastAPI", "MLOps"],
    "Data_Stack": ["Pandas", "NumPy", "SQL", "PostgreSQL", "BigQuery"],
    "Vector_DBs": ["Qdrant", "Pinecone", "Chroma", "FAISS"],
    "Cloud": ["Azure", "GCP", "Docker", "Kubernetes"],
    "Languages": ["Python (Expert)", "SQL (Advanced)", "German (A2.1)", "English (Fluent)"]
}