Beschreibung

Senior AI/ML Engineer and team lead with 7+ years building production ML, LLM, and MLOps systems in finance and SaaS. Leading a 3-engineer team at Proxify on an Azure-based, GDPR-compliant voice-to-structured-output platform; previously built ML pipelines adopted by 30+ data scientists, cutting delivery time by 80%, and a six-model loan pricing ensemble that lifted profit by 87%. Comfortable as both an IC and tech lead in distributed teams.

Branchenexpertise

Sprachen

Türkisch
Muttersprachlich oder zweisprachig
Englisch
Verhandlungssicher
Französisch
Konversationssicher

Arbeitsortpräferenzen

Nur remote

Führt Projekte hauptsächlich remote aus

Proxify
Senior AI/ML Engineer & Team Lead
September 2025 - Heute (9 Monate)
Stockholm-based vetted remote tech talent network serving companies across Europe and the US.

Project 1: Voice-to-Structured-Output SaaS
•Architected and deployed a GDPR-compliant Azure SaaS for voice-to-structured-output, integrating gpt-4o-transcribe and gpt-5 across ACR, Web Apps, and Key Vault, with 200+ production users.
•Led 2 backend engineers and 1 DevOps engineer, coordinating with 2 business stakeholders to deliver the system end-to-end with secure secrets management and a scalable cloud architecture.
•Owned 31 REST endpoints, Celery/Redis async processing, PostgreSQL, and Dockerized CI/CD on Azure from design through production.
•Ran a vendor lock-in and cost analysis benchmarking Qwen3.5-9B (INT8 via vLLM) and whisper-large-v3 co-located on a single T4 against the Azure gpt-5 + gpt-4o-transcribe stack; identified the self-hosting crossover at 1.6k requests/day and quantified throughput at 20 token/s decode.

Project 2: SageMaker Inference Optimization
•Reduced SageMaker inference costs by 93% across 13 NLP inference endpoints, from $10,000 to $600 per month, via model optimization and incremental rollout validated through shadow inference, offline-online accuracy comparison, and staged traffic cutover, maintaining prediction accuracy within 1 pp of baseline.
•Built a Grafana observability stack with CloudWatch and SQL Server dashboards for endpoint health, queue depth, prediction drift, and daily cost tracking across 13 production inference endpoints.
LLM MLOps Microsoft Azure FastAPI SQL
TURING
Senior AI/ML Engineer
SOFTWARE-HERSTELLER
Dezember 2023 - September 2025 (1 Jahr und 9 Monate)
San Francisco-based AI focused talent platform working with frontier labs including OpenAI, Google, and Anthropic.

•Shipped an internal RAG chatbot with Llama 2, sentence-transformers embeddings, FAISS, FastAPI, and Gradio, enabling Q&A over 5,000 company documents and serving 800 queries per week.
•Delivered 3,000+ RLHF evaluations for Python and ML coding tasks at a 98% acceptance rate, contributing reference solutions across Pandas, PyTorch, TensorFlow, Hugging Face, and LangChain to post-training pipelines.
•Reproduced methods from 40+ ML/AI research papers as runnable notebooks across LLMs, computer vision, NLP, and time series, with a 95% acceptance rate.
Python LLM MLOps Machine learning Pytorch
TEB-BNP Paribas
Senior Data Scientist
März 2021 - Dezember 2023 (2 Jahre und 9 Monate)
İstanbul, Türkiye
One of Turkey's largest private banks, part of the BNP Paribas Group.

•Built and maintained a FastAPI-based end-to-end ML pipeline app covering preprocessing, feature engineering, feature selection, hyperparameter optimization, and model evaluation. Adopted by 30+ data scientists across the ML department and used to deploy 30+ production models, cutting project delivery time by over 80%.
•Increased loan profit by 87% by combining six ML models into a retail loan pricing optimization system, validated through a progressive A/B testing rollout tracking profit per loan, utilization rate, and risk profile distribution at each stage before full deployment.
•Increased FX spread revenue by 37% by building a FastAPI-based FX pricing service that used GMM customer segmentation to assign sensitivity scores and generate dynamic spread recommendations.
•Owned production deployment, monitoring, and ETL optimization for multiple models on on-prem infrastructure using Docker, Jenkins, Pytest, Pydantic, Airflow, SQL and ElasticSearch.
FastAPI SQL Python MLOps Pytorch