Serkan UYSAL

MACHINE LEARNING ENGINEER

Sahibinden.com•  April 2022 - Present•  Istanbul,TR

Used various open-source and commercial LLM models combined with tools such as LangChain, ElasticSearch, etc., and deployed on FastAPI in internal projects. Transformed the deep learning price recommendation model to the dockerized automated version with statistical methods and automated cron jobs (increased accuracy by 10%). Created a Grafana panel due to the model performance monitoring. Collected and preprocessed car photos, trained a car parts detection model with PyTorch on GCP, and served the model as a Python package for use. Reached 90% accuracy on 17 classes. Trained a computer vision model on TensorFlow that classifies car models with a single image or multiple images. Converted a sync Flask service to an async Kafka service.

DATA SCIENTIST

Wope•  October 2021 - March 2022•  Remote

Focused on SERP (Google Search) data and big data, using title, description, and meta-data. Found similar titles on SERP by using FastText. Developed an application that visualizes similar titles using Dash framework on a cloud provider.

MACHINE LEARNING INTERN

Robogör•  January 2021 - March 2021•  Istanbul,TR

Analyzed Semantic Segmentation architectures with manipulated aerial images to determine a safe landing location. Trained a DQN (Reinforcement Learning) model with AirSim that attempted to optimize pitch, yaw, and roll rates to reach a safe landing location as quickly as possible.

MACHINE LEARNING INTERN

Baykar Technology•  January 2018 - April 2018•  Istanbul,TR

Conducted academic research about UAV and its components. Built a monitoring application with Tkinter and Matplotlib that can visualize the streaming input.