Mohammad-Reza Azizi

5 roles · 2020 — 2026

Experience.

LLM applications, recommendation systems and production Python services, from speech models to RAG chatbots.

01 Roles

  1. Mar 2025 — Feb 2026

    MyTrek AI

    Senior AI Engineer

    • Built a content-based video recommendation system using LLM-generated embeddings, Qdrant, and KNN, with automated data pipelines in Apache Airflow and APIs built with FastAPI and Redis caching.
    • Developed an awareness detection module for a KYC workflow using MediaPipe head-pose estimation, helping improve the reliability of user verification.
    • Built production Python microservices with PostgreSQL, Redis, gRPC, and Apache Kafka, supporting reliable service-to-service communication across the platform.

  2. Mar 2024 — Mar 2025

    Ceres Technology

    AI Engineer

    • Built and trained LightGBM models on supply-chain data to estimate delay risk, and turned the results into clear charts that helped business teams understand patterns and make better decisions.
    • Developed a supply-chain analytics chatbot using LangChain, Vanna.ai, OpenAI, and Plotly, allowing users to upload files, query external sources, combine datasets, generate visualizations, and retrieve relevant news.
    • Built a dataset recommendation service that compared user data against more than 300k datasets and returned relevant results in under 15 minutes using efficient multiprocessing.

  3. Mar 2023 — Mar 2024

    Core Decentralized Technologies

    AI Engineer

    • Built a content-based video recommendation system using LLM-generated embeddings, storing vectors in Qdrant and finding related videos with KNN. Automated the pipeline with Apache Airflow and exposed it through FastAPI with Redis caching.
    • Developed an awareness detection module for a KYC workflow using MediaPipe head-pose estimation, improving the accuracy of user verification.
    • Expanded into broader software engineering by creating Python microservices with PostgreSQL and Redis, implementing service-to-service communication with gRPC and Apache Kafka, managing database migrations with Alembic, and writing thorough unit and integration tests.

  4. Feb 2021 — Mar 2023

    Digikala

    Data Scientist

    • Built a complementary recommendation system using KNN on image and text embeddings stored in Qdrant, increasing add-to-cart rate by 70% and daily GMV by 202%, validated through a diff-in-diff experiment.
    • Optimized a trend-finding algorithm by adding parallel processing across the product catalog, achieving a 1200% speedup in offline processing tasks.
    • Processed large-scale datasets with Apache Spark and scheduled recurring offline workflows using Apache Airflow.

  5. Jan 2020 — Feb 2021

    Part AI Research Center

    Data Scientist / Speech Chapter Lead

    • Trained and evaluated speech models using TensorFlow, PyTorch, ESPNet, Nvidia NeMo, and Optuna, while leading a team of four data scientists through task planning, code reviews, and daily coordination.
    • Researched and experimented with state-of-the-art ASR models in English, testing different architectures and training approaches.
    • Created and maintained text and speech datasets, including data collection, cleaning, denoising, and quality control.

Full details in my résumé (PDF).