ParhamKhoshSolat

For three years I was the data person at Sakoudar, an influencer marketing platform. The problem underneath it was comparison: raw impressions reward whoever posts most often, and a beauty audience is nothing like a tech audience, so nobody could say which creator was actually worth the money. I built the weekly pipeline that pulled engagement from the Instagram, LinkedIn and X APIs, the scoring that corrected for both distortions, and the recommendation engine that turned a customer's brief into a ranked shortlist. The platform is still running.

In April 2024 I stood in front of Fater S.p.A.'s leadership (the Procter & Gamble joint venture behind Pampers Italia) presenting a map that ranked twenty Naples districts by store-expansion potential. Behind the map: MySQL joins between Fater's proprietary sales records and Italian census data, plus a Python geospatial pipeline. In front of it: an executive who could read the answer in five seconds. The jury recognised the project.

That pattern repeats across this site. A vision-language model I fine-tuned on 60,000 image-question pairs, shipped as a public web demo where anyone uploads a photo and asks questions about it. A Kafka and Zookeeper cluster that streamed six months of price data through producers, with a separate PySpark stage that clusters a ticker's price history. A tournament of eight model families that chose the pest-outbreak detector that caught every outbreak day in the held-out set, at the cost of a false alarm rate a farmer can live with.

My newest work leans into reinforcement learning. Six warehouse vehicles routed by classical and learned planners alike, all under a collision shield that makes crashes structurally impossible. An agent that repairs damaged photos by chaining small specialist networks, reimplemented from a CVPR paper and extended with its unreleased fine-tuning. Both are live demos you can try. The same thread runs into my MSc thesis at Federico II, on how a learning robot can stay transparent and trustworthy to the person it learns from.

What I'm after: a team that values clarity over cleverness and ships work to real users. Strong in Python, SQL (CTEs, window functions, query optimisation), PyTorch, HuggingFace Transformers, reinforcement learning, Apache Kafka, PySpark, and Power BI.

Naples-based, with three years of paid data science behind me and an MSc finishing this year. Data Scientist, ML Engineer, or AI Engineer roles in Italy, Germany, or remote across the EU. English C1.