01
Candidate understanding
Extract structured candidate information from CV data and preferences.
Machine Learning Engineer
Umut Gökdemir
NLP · Agentic AI · Applied Machine Learning
I build and evaluate machine learning systems, from transformer-based NLP models to semantic retrieval pipelines and agentic AI workflows.
Featured work
Agentic AI · ML Systems
CareerMatch Agent
An explainable AI-assisted job-matching system that turns a candidate profile and job preferences into ranked recommendations using deterministic constraints, multilingual semantic retrieval and evidence-grounded LLM evaluation.
Python
LangGraph
FastAPI
SentenceTransformers
OpenAI
Gemini
Ollama
System workflow
01 → 0402
Hybrid retrieval
Combine deterministic constraints with multilingual semantic ranking.
03
LLM evaluation
Evaluate matches with evidence-grounded reasoning instead of opaque scoring.
04
Explainable output
Return ranked opportunities with clear supporting evidence and rationale.
Multi-provider architecture
Interchangeable OpenAI, Gemini and Ollama model providers.
Evaluation-driven
Benchmark tooling for ranking quality, retrieval metrics and latency.
Engineering-first
API-based design with automated tests and reproducible workflows.
Selected projects
Applied ML across language and scientific data
A small set of projects chosen for technical relevance, measurable results and clear evidence of hands-on machine learning work.
Natural Language Processing
Multilingual Social-Media NER
Fine-tuned transformer-based models for Turkish named entity recognition on noisy social-media text, supported by improved annotation and preprocessing workflows.
Improved span-level micro-F1 from 0.44 to 0.75.
PyTorch
Hugging Face
Transformers
Model Evaluation
Applied Machine Learning
Predictive Maintenance ML
Built an end-to-end predictive-maintenance pipeline for imbalanced equipment-failure classification, combining domain-informed feature engineering, cross-validated model selection, threshold optimization and systematic error analysis.
Achieved 0.876 held-out test Average Precision with 0.947 precision and 0.794 recall.
Python
scikit-learn
XGBoost
Model Evaluation
Technical toolkit
Tools I use to build, evaluate and engineer ML systems
My strongest work sits at the intersection of machine learning, NLP, AI systems and production-minded Python engineering.
Machine Learning
PyTorch
scikit-learn
pandas
NumPy
Deep Learning
Model Evaluation
Clustering
NLP & Retrieval
Hugging Face
Transformers
SentenceTransformers
Semantic Search
Embeddings
Named Entity Recognition
AI & LLM Systems
LangGraph
Agentic Workflows
LLM Evaluation
Evidence-Grounded AI
OpenAI
Gemini
Ollama
Backend & Applications
Python
FastAPI
Pydantic
REST APIs
HTTPX
Streamlit
Testing & Code Quality
pytest
Unit Testing
Integration Testing
mypy
Ruff
Type Checking
Infrastructure & Development
Docker
Git
GitHub Actions
CI/CD
Linux
AWS
Professional experience
Machine Learning Experience
Applied machine learning experience across NLP and high-dimensional scientific data, spanning data preparation, model development, evaluation and reproducible experimentation.
Machine Learning Engineer Intern
EnlightyAI
May 2025 – October 2025
Remote
- Improved span-level micro-F1 from 0.44 to 0.75 by fine-tuning and evaluating transformer-based NER models for noisy Turkish social-media text.
- Improved training-data quality across 6,000 annotated tweets by coordinating annotation and validation workflows, refining preprocessing and systematically investigating annotation and prediction errors.
- Strengthened model selection and experimentation by benchmarking multiple BERT and XLM-R variants through reproducible training, validation and error analysis.
- Led an 8-person internship team by planning workloads, assigning tasks, running weekly progress meetings and consolidating progress reports, helping keep annotation, preprocessing, training and evaluation work aligned across the project.
Machine Learning Engineer Intern
Functional Genomics Laboratory Regensburg
March 2024 – June 2024
Regensburg, Germany
- Reduced the analysis data footprint by approximately 40% by building a Python preprocessing pipeline for cleaning, restructuring and optimizing high-dimensional DESI mass-spectrometry imaging data.
- Preserved 95% of explained variance while reducing feature dimensionality by applying PCA before unsupervised tissue-region segmentation with four-cluster K-means.
- Improved the interpretability of unsupervised tissue segmentation by mapping cluster assignments back to spatial coordinates and validating molecular-density patterns across follow-up scan series.
- Worked within an interdisciplinary research team by communicating analysis progress through weekly reports, incorporating domain feedback into the workflow, presenting the methodology and results to the team and stakeholders.
Leadership & community
Leadership Beyond Engineering
Long-term experience in team leadership, mentoring, operations and community organization.
Scout Leader & Association President
Camadan Scout, Youth and Sport Association
April 2014 – April 2021
- Led the training and development of approximately 15–20 scouts each year by designing recurring educational programs, preparing training materials and delivering practical and theoretical sessions.
- Managed camps and community activities end to end, coordinating logistics, schedules, equipment, accommodation, safety and communication with participants and families.
- Directed association operations as President through volunteer coordination, delegation, budgeting, external partnerships and leadership meetings, while mentoring senior scouts into future leader and instructor roles.
Research & education
Quantitative foundations for machine learning
M.Sc. | 2026
Computational Science
University of Regensburg
Advanced training in machine learning, numerical methods, scientific computing and computational modelling, with hands-on work on data-driven and simulation-based problems.
B.Sc. | 2021
Physics
Middle East Technical University
Built a rigorous foundation in mathematical modelling, numerical problem solving and computational physics, including research experience in quantum many-body systems.
About
Physics-trained. Engineering-focused.
I am a Machine Learning Engineer with a background in computational science and physics. I build and evaluate ML systems across NLP, semantic retrieval, agentic AI and scientific data, with an emphasis on reproducible experimentation, measurable performance and systematic error analysis. My quantitative background shapes how I approach modelling and uncertainty; my engineering focus is turning those ideas into reliable, testable systems.
Contact
Open to machine learning opportunities
I am interested in Machine Learning Engineering, Applied ML and AI Engineering opportunities where I can work on meaningful models, data and production-oriented AI systems. For opportunities, collaborations or technical conversations, feel free to reach out.