AI Developer

Onur  is an experienced AI and Machine Learning Engineer with almost seven years of professional experience, mainly focused on Natural Language Processing (NLP), Large Language Models (LLMs), and data-driven AI systems. After completing his studies in Korea, Onur began his career in 2019 and has since contributed to various end-to-end AI projects, from model design and fine-tuning to deployment and optimization. Throughout his career, he has worked extensively on NLP-based applications, including search and recommendation engines, question-answering systems, chatbots, and document retrieval. In recent years, his focus has been on LLM development and RAG (Retrieval-Augmented Generation) architectures, where he has built and fine-tuned large language models (including for Turkish), integrated them into real-world use cases such as customer support chatbots, and implemented information extraction and code conversion systems. Onur is highly skilled in Python as his main programming language and has a working understanding of Java. His technical toolkit includes Milvus, Pinecone, Chroma, and Elasticsearch for vector databases, alongside Docker and Kubernetes (AKS) for deployment. He also has solid hands-on experience with Azure and Databricks, and basic exposure to AWS. Additionally, he has used LangGraph and CrewAI for agentic workflows and is comfortable with Jira and Azure DevOps for project management

Bachelor — Kangnam University

Computer Science

Senior Machine Learning Engineer — Sep 2024 – Present

• Led encryption/decryption of data in AWS S3 bucket for enhanced security measures.

• Conducted data migration from MongoDB to AWS using multiprocessing for improved efficiency.

• LLM Code Conversion module to convert data processing packages from Java to Snowpark Python that is integrated fully in the Snowflake environment using the llama model through Snowflake Cortex.

Principal Specialist, Artificial Intelligence Solution Management | Etiya — Jul 2024 – Sep 2024

• Led a ML team in predicting course completion rates and recommending new courses for course participants.

• Deployed predictive model through Flask API for seamless integration and user experience

Data Scientist | Nesine.com — Aug 2023 – Apr 2024

• Developed QA models on GPT models for Nesine.com, enhancing customer service and help desk support.

• Utilized DeepSpeed for distributed LLM model training on multiple GPUs, improving efficiency.

• Implemented RAG application as a help desk chatbot for new users, utilizing Elasticsearch and OpenAI gpt-4-turbo model.

NLP Machine Learning Engineer | EY — May 2023 – Aug 2023

• Developed a chunk-based Semantic Grading System for EY ConvoCoach platform using word embeddings for answer similarity.

• Implemented a Data Augmentation Rag pipeline to generate synonyms for natural language regular expressions.

• Deployed the system on AKS for efficient performance and scalability

Research Engineer | Huawei — Apr 2021 – May 2023

• Working on user intent extraction from search queries on Huawei App Gallery.

• Clustering queries in Turkish, Russian, and other required languages to annotate them, train a classification model to extract information on which categories users are searching on to display better results.

• Participated in a research paper project on topic-based text clustering. Experiment with different clustering algorithms on different datasets for the evaluation and to set a benchmark.

Specialist, Artificial Intelligence Solution Management | Etiya — Oct 2020 – Mar 2021

• Completed and put in the production of a semantic search engine particularly for e-commerce websites when searching for products.

• In this project, I used Logstash to index the data into Elasticsearch. Created sentence vectors for the product item names using fastText pre-trained word embedding model and indexed them in Elasticsearch.

• By utilizing the cosine similarity feature from Elasticsearch, I obtained the semantic search results.

• In addition to that, I created a personal recommendation section through the user shopping log data, created vectors for all products a user had bought, and which brings the recommended products for that particular user to the homepage.

• Training keyword extraction model using BERT on AWS server.

Senior Research Engineer | Xinapse — Mar 2019 – Mar 2020

• BERT Machine Reading Comprehension(Question&Answering), Fine-tuning, optimization, SentencePiece tokenization implementation for English and Korean

• BERT Sentiment Analysis

• BERT Text Classification with TREC-6 dataset

• Time Series Forecasting of Business Cycle Indexes (News data-based) (Korea Institute of Industrial Technology)

• Statistical Data Analysis using Unstructured Dataset

• Topic Modeling, LDA, NMF, Dynamic Factor Modeling

Machine Learning Engineer | Choongwae Information Technology — Oct 2018 – Feb 2019

• RNN Sequence-to-Sequence model based patient service system chatbot for a university hospital (Electronics and Telecommunications Research Institute Korea - ETRI).

• Seq-to-Seq model built-in raspberry pi

• Database Management, Data Migration, Data Manipulation & Parsing & Extraction from the database (Oracle, SQL).