Professional Summary
AI Engineer specializing in production LLM systems, agentic RAG pipelines, and NLP applications. Built and deployed LangGraph-powered agents with adaptive retrieval, hallucination detection, and self-correcting workflows. Delivered AI platforms used by 50+ researchers, reducing manual processing time by 60%. Experienced across the full ML lifecycle — from vector store design and LLM orchestration to model fine-tuning on Hugging Face, FastAPI development, GDPR-aware data handling, and cloud deployment.
Technical Skills
Languages & Frameworks
PythonJavaScriptDjangoNext.jsReactHTMLCSSBootstrapTailwind CSS
Databases
PostgreSQLMySQLSupabaseChromaDB
Tools & DevOps
GitDockerHerokuFastAPIVS Code
AI/ML
CNNTensorFlowScikit-LearnNLPLLMRAGNERFine-tuning
AI Systems & APIs
LangGraphLangChainHugging FaceOpenAIClaudeOllama
AI-assisted Development
GitHub CopilotCursor AI
Languages
English (C1)German (B1)Persian (Native)
Experience
Applied AI Developer
University of Freiburg, Freiburg, Germany
- ✦Designed and deployed LLM-powered applications for automatic semantic metadata generation used by 50+ researchers
- ✦Built AI pipelines combining NLP, validation logic, and human-in-the-loop workflows for annotating texts, reduced processing time by approximately 60%
- ✦Integrated AI services with backend systems and databases for production use
- ✦Collaborated with interdisciplinary teams to implement semantic web standards (RDF, OWL, SHACL) for knowledge graph construction
Scientific Researcher
Fraunhofer IWM, Freiburg, Germany
- ✦Built Python scripts and automation tools for large-scale computational workflows
- ✦Developed AI-supported data analysis pipelines using TensorFlow and Scikit-Learn
- ✦Improved simulation workflow efficiency by 80% through algorithmic optimization
- ✦Published 2 peer-reviewed papers and supervised 3 junior researchers and master's students
Postdoctoral Researcher
Uppsala University, Uppsala, Sweden
- ✦Designed and maintained Python tools for processing scientific datasets
- ✦Automated data analysis, visualization, and reporting pipelines
- ✦Supported collaborative research through reusable Python modules
- ✦Established international collaborations resulting in 2 joint research proposals
Education
Diploma: Full Stack Software Development
Code Institute – Dublin, Ireland
Ph.D.: Nanotechnology Engineering
Kashan University - Kashan, Iran
MSc & BSc: Physics
Ferdowsi University - Mashhad, Iran
Selected Projects
EU AI Act Compliance Intelligence RAG Agent
2026LangGraphChromaDBFastAPIStreamlitHeroku
- •Adaptive RAG agent built with LangGraph and ChromaDB for querying the EU AI Act
- •Implements corrective retrieval pipeline: document grading, hallucination detection, and web search fallback
- •Exposed via FastAPI REST API with a Streamlit demo; deployed on Heroku Cloud
Fine-tuned MatSciBERT on Chemical Dataset
2025MatSciBERTHugging FaceNERFine-tuningPython
- •Fine-tuned m3rg-iitd/matscibert on the CHEMDNER corpus (19,440 annotated biomedical examples) for chemical entity recognition
- •Achieved F1 of 0.91 on chemical entity spans using Hugging Face Trainer API with BIO tagging and seqeval evaluation
LLM-Powered Annotation Application
2025StreamlitPythonOpenAIClaude AINERLLM
- •Automatically annotates scientific text with LLMs
- •Chunk-based processing with multi-format export
- •Improved annotation efficiency by 70%
ML Plant Disease Classification
2023TensorFlowCNNStreamlitPythonMLOps
- •CNN model achieving 95% accuracy on plant disease detection
- •Optimized inference time and integrated into a full-stack web application with FastAPI and Next.js
- •Leveraged an LLM to generate treatment recommendations and disease explanations for detected plant conditions
Metadata Schema Generator with LLMs
2025Django 4.2PostgreSQLOllama AISupabaseBootstrap
- •Auto-generates research metadata using LLMs
- •User authentication, file upload, and real-time processing
- •Reduced metadata creation time from 2 hours to 10 minutes
RDF/SHACL Generator with AI Agents
2024PythonFastAPIRDFLibNetworkX
- •Converts material test data into structured knowledge graphs
- •SHACL validation ensuring 95% data quality compliance
- •Integrated with existing research infrastructure
NextJS + FastAPI Full-stack RDF/SHACL Application
2025FastAPINextJSAI ChatboxTailWind CSSRDFSHACL
- •Workflow for evaluating experimental/modeling methods and uncertainties
- •Deployed on Heroku and Vercel with CI/CD pipeline
- •Implements community-driven processes for reference material data sets
E-Learning Booking Platform
2024DjangoPostgreSQLBootstrapSupabase
- •Full CRUD operations for courses and bookings
- •Email notifications and administrative dashboard
- •Responsive design for mobile and desktop