Skip to content

Resume

Download Resume

ABDULLAH SAMIL GUSER

abdullahsamilguser@gmail.com | abdullahsamilguser.com

ML Engineer with 9 years of experience applying deep learning, NLP, and LLMs to turn text-heavy data into customer-facing products. Product-minded and iterative, I design offline evaluations, run online experiments, and ship production ML pipelines that measurably improve user outcomes.

WORK EXPERIENCE

SVITLA - Senior Machine Learning Engineer (January 2026 – Present; Full-time & Contract)

Tech Stack : PyTorch, Transformers, HuggingFace, Llama, Qwen, RunAI, Nebius, W&B, Google Cloud (GCP), Github, PEFT, LoRa

  • Small Language Model (SML) SFT Fine-tuning: Fine-tuning 1B & 0.5B parameter on-device models for domain-specific applications. Use PEFT methods like LoRa to reduce resource requirements while maintaining performance.
  • Teacher-Student Knowledge Distillation: Distill 3B & 7B parameter teacher models into 0.5B & 1B student models. Experiment with different distillation techniques (on-policy, off-policy, forward KL, reverse KL etc.)
  • W&B Reports: Create detailed W&B reports to analyze training runs, compare models, and visualize performance. Create W&B Sweeps to automate hyperparameter tuning and find optimal training configurations.

LEADIQ - Machine Learning Engineer (January 2022 – December 2025; Full-time & Contract)

Tech Stack : PyTorch, Transformers, HuggingFace, GPT/OpenAI API, LangChain, FastAPI, GraphQL, Streamlit, MLflow, Databricks, PySpark, Redshift, AWS, Docker, Kubernetes, Datadog, Redis, MongoDB, GitLab, Terraform, S3, Lambdas, ArgoCD, CI/CD, Kafka

  • Experimentation & Evaluation: Set up online and offline tests and experiments to compare LLM prompts/models across cohorts; using latency, quality, and cost metrics to gate releases and prevent regressions.
  • Personalization with LLMs: Delivered LLM-powered email generation, owning data curation, prompt lifecycle, evaluation dashboards, and alerting for reliability at scale.
  • RAG & Chatbot: Built embeddings-based chat and Q\&A with standardized retrieval and prompt templates, plus fallbacks to keep answer quality consistent across heterogeneous documents.
  • Agent Workflows: Delivered agents and tools that research target accounts and produce structured outputs; scheduled pipeline runs and integrated results with downstream systems via APIs & data tables.
  • Applied ML: Trained title seniority/function classifiers, a Siamese title similarity model, an XGBoost revenue forecasting model etc.
  • APIs & Delivery: Exposed capabilities via FastAPI and GraphQL, automated CI/CD to Kubernetes and Databricks, added monitoring and incident guides with SLOs, and enforced type-safe Python interfaces.
  • Product-Focused ML: Worked with teammates to clarify objectives, define measurable success criteria, and prioritize prototypes that could ship safely behind flags and iterate based on real usage.

SESTEK - Senior AI Research Engineer (March 2020 – December 2021; Istanbul, Turkey)

Tech Stack : PyTorch, BERT, LSTM, FastAPI, Transformers, HuggingFace, Docker, Azure (deployment), Keras, spaCy, NLTK, Scikit-Learn, StanfordNLP, FastText

  • Speech & Sequential Modeling: Trained CNN/RNN/LSTM/BERT-based models for emotion, age, gender, and spoofing detection; chatbot intent detection, named entity recognition, text summarization, text correction etc. all of which are used in production. Exposed models as REST services with FastAPI and Docker.
  • Mentorship: Mentored junior researchers/engineers to improve onboarding speed and research-to-production efficiency.
  • Research leadership: Conducted literature reviews, wrote reports, and supported audits to secure and extend R\&D funding.

SESTEK - AI Research Engineer (March 2017 – March 2020; Istanbul, Turkey)

  • Intent modeling & NLP: Built & deployed on-premise chatbot intent detection (LSTMs, BERT, Hugging Face) models for large banks.
  • Seq2Seq modeling: Developed an LSTM sequence-to-sequence model for normalizing textual numeric expressions (addresses, dates, times); published at IEEE SIU 2019.
  • Full ML pipeline ownership: Data cleaning, training, evaluation, testing on real customer data, deployment as REST services.

EDUCATION

BOGAZICI UNIVERSITY (Istanbul, Turkey)

Master of Science: Electrical and Electronics Engineering (2017-2020) (GPA: 3.69/4.00)

  • Research Focus: Automated Response Generation for Corporate Chatbot Systems
  • Specialized Courses: Pattern Recognition, Speech Processing, ML, Statistical Signal Analysis, Social Semantic Web

BOGAZICI UNIVERSITY (Istanbul, Turkey)

Bachelor of Science: Electrical and Electronics Engineering (2013-2017) (GPA: 3.28/4.00)

  • Core Courses: Probability, Matrix Theory, Signal Processing, Introduction to Image Processing

TECHNOLOGIES

  • Languages/Frameworks: Python, PyTorch, Transformers, HuggingFace, LangChain, FastAPI, GraphQL, Pydantic/mypy, OpenAI API, Strawberry, Streamlit, LlamaIndex, MLflow, Keras, PEFT/LoRa, C#, Javascript
  • ML/AI: Chatbots, Recommender Systems, Ranking, Personalization, LLMs, SLM Fine-tuning, Transformers (BERT, GPT, Llama, Qwen), LSTMs, Siamese Networks, Gradient Boosting, FAISS, ONNX, Kaldi, Generative AI
  • Data & Infra: Databricks, PySpark, Redshift, SQL, PostgreSQL, AWS, Google Cloud (GCP), S3, Lambdas, Docker, Kubernetes, Terraform, RunAI, Nebius, W&B, GitLab CI/CD, ArgoCD, Datadog, Redis, MongoDB, Azure, Kafka, Salesforce, Microservices
  • Testing & Tooling: Postman, API Testing, Unit Testing, PyTest, End-to-End Testing
  • CI/CD: GitLab, GitHub, Jenkins, Slack, Jira

PUBLICATIONS

  • A. S. Güser, M. Erden and M. L. Arslan, "Semi-Automatic Formatting of Spelled Out Numbers" 2019 27th Signal Processing and Communications Applications Conference (SIU), Sivas, Turkey, 2019

CERTIFICATES

ADDITIONAL

  • Achievements : Ranked 6th among 1.7m students in the National University Entrance Exam (2012)
  • Activities : Walking, Reading, Traveling