Data Scientist - GARANTI BBVA
Hi, I’m Taner Sekmen 👋
I am a Data Scientist and Software Developer based in Istanbul, Turkey, with a strong interest in open-source development, machine learning, and software engineering.
I am currently developing projects at Garanti BBVA while actively contributing to open-source communities and company repositories, including
Microsoft,
Google Research,
Hugging Face,
GitHub Education,
EthicalML, and
Data Talks Club.
I have opened issues and explored opportunities to support the development of projects such as
OpenAI,
Obsei, and
NetworkX.
Additionally, I made a direct contribution to
LangChain
by improving and expanding example documentation.
Beyond code contributions, I translated and deployed the Turkish version of the
promptingguide.ai
website, a platform focused on artificial intelligence, natural language processing, and prompt engineering, which has over 20,000 followers.
Most recently, I contributed to the
awesome-chatgpt-prompts
repository (100,000+ stars), with a focus on enhancing prompt input format characteristics.
- Financial summaries of PDF and TXT documents are generated using Llama and Qwen language models.
- The architecture of language models in GGUF format was researched and an application was developed.
- LayoutLM was utilized for information extraction from documents.
- Integration of the Docling VLM framework was performed on the server.
- A project was developed using the Gemma language model to correctly classify input text and additionally return a generic response.
- Summarization of incoming customer call data was achieved using the Gemma language model.
- By applying the agentic AI approach, certain processes were automated.
- Documents received from the relevant departments were processed through OCR and converted into text. Subsequently, an open-source LLM was utilized to develop a QA chatbot for the operations teams.
- Developed solutions for ad-hoc requests.
- Built a RAG system that operates in a text-to-SQL format.
- Designed the backend of the developed chatbot using Flask, created a Docker image, and deployed it on Cloud Run.
- Applied Association Rule Learning to identify hidden patterns and improve recommendation systems, enhancing customer experience and driving engagement.
- Developed a Machine Learning-based Activation Model to optimize user engagement and retention through predictive analytics.
- Generated customized reports, conducted in-depth data analysis, and tailored existing studies to align with specific organizational requirements.
- Prediction of customers' pin location over 10M unique client using machine learning algorithms e.g. Gaussian Mixture Model, Hierarchical-based, K-Means, BIRCH
- A cost reduction of over 500k $ was achieved through the pin location project by annually.
- Worked with AWS Redshift, PostgreSQL, and Athena database systems.
- Built reports and cron jobs.
- The ad filtering system with natural language processing methods for GetirJobs has been enhanced using the BERT model
- Developed topic modeling using the natural language processing method BERTopic based on the feedback obtained from couriers.
- Created a baseline NLP model to detect profanity.
- Utilized APIs to write data into our databases from third-party services.
- Developed interactive dashboards utilizing Qlik Sense to enhance data visualization and facilitate insightful business analysis
- Examined shop and delivery data sourced from SAP, implementing comprehensive data validation processes to ensure accuracy and reliability
- Conducted in-depth analysis of missing data, identifying optimization opportunities and implementing strategic solutions to enhance overall data quality
- Coded object oriented programming to optimize customer portfolio
- Worked a financial theory to invest
In the dataset, It is the 2nd most star-rated notebook among the notebooks studied for this dataset.