I'm Bahador Mirzazadeh, a Data Scientist and Full-Stack AI Engineer based in Milan, Italy, with 6+ years of software engineering behind me. I hold an M.Sc. in Data Science from the University of Padua and a B.Sc. in Computer Science from Tabriz University.
I work where machine learning meets real systems — training the model and shipping the service that serves it: distributed asynchronous processing pipelines, high-compute medical imaging, real-time computer vision, and LLM agents that talk to production databases.
class Bahador:
role = "Data Scientist / Full-Stack AI Engineer"
location = "Milan, Italy 🇮🇹"
focus = ["Deep Learning", "Computer Vision", "LLM Agents", "MLOps"]
stack = ["Python", "PyTorch", "TensorFlow", "LangGraph", "FastAPI", "Django"]
languages = ["Persian (native)", "English (C1)", "Italian (A2)"]
open_to = ["research collaborations", "AI engineering roles", "good coffee ☕"]🔬 Interests: deep learning architectures · image colorization & generative models · multimodal and audio ML · retrieval and agentic LLM systems · MLOps / LLMOps
🌱 Currently exploring: distributed training at scale, agent orchestration patterns, and cloud-native LLM deployment.
📫 Reach out if you want to talk research, AI engineering, or an interesting problem — I'm always up for it.
Languages
Machine Learning & Deep Learning
LLMs & Agents
Backend & Infrastructure
Data Stores
| Project | What it is | Stack |
|---|---|---|
| DB-AI-Agent | A LangGraph agent that turns natural language into database operations, served behind an API. | LangGraph FastAPI PostgreSQL |
| iColoriT_TF | TensorFlow implementation of iColoriT — ViT-based interactive image colorization, with both sinusoidal and learnable positional embeddings. | TensorFlow Vision Transformer |
| Vision_Transformer_TF | ViT built from scratch, following An Image is Worth 16x16 Words. | TensorFlow Keras |
| Environment_Sound_Detection | Environmental sound classification on ESC, comparing InceptionV4, DenseNet121 and a stacked-CNN ensemble. | TensorFlow Audio ML |
| Human-Voice-Predictor | Real-time gender and age recognition from audio using a CRNN. | CRNN Audio ML |
| Weather-Classification | Custom CNN for weather classification across four merged datasets. | TensorFlow CNN |
| Chest-Xray | Deep learning on chest X-ray imaging (Human Data Analytics). | TensorFlow Medical Imaging |
| Nonlinear_Optimization_Project | Iterative methods for finding optimal points of functions in ℝⁿ. | Python Optimization |
| encryption | The BlackPlatinum encryption component for PHP. | PHP Cryptography |
M.Sc. Data Science — University of Padua, Italy (2022 – 2025)
Thesis: Disagreement-Aware Multi-Teacher Distillation for Perceptual Image Colorization — using disagreement between teacher models as a signal to guide knowledge distillation toward perceptually better colorization.
B.Sc. Computer Science — Tabriz University, Iran (2016 – 2021)
- 🧩 I like algorithmic puzzles more than I like admitting how long I spend on them.
- 📄 I read ML papers the way other people read the news — then try to reimplement them from scratch.
- 🎨 My thesis was about teaching models to color grayscale images the way a human sees it, not just the way a loss function likes it.
- 🌍 Three languages, three countries, one editor config.
Thanks for stopping by — if you're building something interesting, let's talk. 🤝