I sit at the intersection of AI engineering and data infrastructure — deploying LLM-powered retrieval systems, architecting modern data stacks, and translating raw transactional data into executive-ready insight.

• 🤖 RAG & LLM Systems — Production Persian chatbot: +38% retrieval precision, ~60% hallucination reduction (FastAPI · Qdrant · OpenAI API)
• 🏗️ Data Engineering — End-to-end ELT pipelines with orchestration, transformation, and quality contracts (Airflow · dbt · PostgreSQL · Docker)
• 📊 Analytics & ML — Churn prediction, RFM segmentation, CLV modelling → Power BI dashboards for stakeholders

Languages & Querying

Data Engineering

AI · LLM · NLP

Analytics & ML

Context-aware document retrieval system deployed for internal operations and external end-users.

Built a full-stack RAG pipeline over a mixed Persian/English document corpus. Implemented a retrieval evaluation framework (Precision@K, LLM-as-judge faithfulness scoring) across 1,000+ documents and 150 domain-specific queries.

Results: Retrieval precision 61% → 84% (+38%) · Hallucination rate reduced ~60%

FastAPI Qdrant OpenAI API OCR Docker Python Sentence Transformers Multilingual Embeddings

End-to-end streaming architecture for real-time financial transaction monitoring and fraud alerting.

Built a decoupled Kafka → Spark → PostgreSQL pipeline capable of handling high-throughput transaction streams. Implemented a tier-one deterministic rule engine with micro-batch processing, secure credential injection via environment variables, and a JDBC sink writing flagged records to an auditable fraud_alerts table. Containerized full infrastructure with single-command spin-up via Docker Compose.

Apache Spark Apache Kafka Confluent Platform PostgreSQL PySpark Docker Docker Compose Python

Production-grade ELT pipeline with automated orchestration and data quality contracts.

Architected a full modern data stack on the Olist e-commerce dataset. Modular dbt project with staging + mart layers, schema tests (not_null, unique, accepted_values), and source freshness checks. Single-command environment spin-up via Docker Compose.

Apache Airflow dbt PostgreSQL Docker Git

End-to-end customer analytics on 100K+ real orders across 9 relational tables (2016–2018).

Built a Python ETL pipeline into PostgreSQL, RFM segmentation framework (Champions / At-Risk / Lost cohorts), Logistic Regression churn classifier (Precision: 0.80, Recall: 1.00), and CLV scoring — all surfaced through four cross-filtered Power BI dashboards.

Python PostgreSQL scikit-learn SQLAlchemy Power BI DAX

Published open-source dataset capturing demographics, purchasing patterns, device usage, and satisfaction metrics — designed for ML, analytics, and educational use.

Python Pandas Open Source

• 🎓 M.Sc. in IT Management (E-Business) · Shahid Beheshti University, Tehran
• 💼 AI Developer @ Barmana — building LLM retrieval systems in production
• 🔍 Open to Data Engineering, AI/ML Engineering, and Analytics Engineering opportunities