Şeref Recep Keskin

Senior AI Engineer

Ankara, Türkiye

About

I bridge the gap between cutting-edge research and production-grade AI systems. I'm an engineer first. I write clean, maintainable code, build robust APIs, and deploy models that handle real traffic. But I also bring a research mindset: 11 published papers and the ability to read a paper, understand the math, and implement it from scratch when needed. My journey spans Defense, HR Tech, Logistics, and Travel, shipping systems that are both intelligent and scalable. Core Competencies: 🔹 Production Engineering: Clean Python, FastAPI, Docker, CI/CD, async systems 🔹 Agentic AI & Generative AI & LLMs: Multi-agent architectures, RAG pipelines, prompt engineering 🔹 Classical ML: Predictive modeling, classification, NLP, and anomaly detection from research to deployment 🔹 Research to Production: Reading papers, prototyping, and shipping I build AI that works, scales, and survives production.

Experience

  • Senior AI Engineer at Prometa AI
    May 2025 - Present · 1 yr 2 mos

    - Leading the development of an autonomous AI travel assistant that handles end-to-end corporate travel planning, reducing hours of manual work to minutes through AI-powered automation. - Designed an event-driven multi-agent architecture where specialized agents collaborate on complex tasks, handling nuanced requests like "find me the cheapest business class flight arriving before my 9 AM meeting" or "find a 5-star hotel in Berlin with a gym and free breakfast." - Added a conversational voice layer to the travel assistant for hands-free trip planning, initially using OpenAI Whisper for STT and later migrating to GPT-4o mini Transcribe for improved latency and accuracy, paired with ElevenLabs for TTS. - Built a document intelligence pipeline extracting financial tables from unstructured documents using PaddleOCR and vision-language models. - Delivered an analytics dashboard for an investment firm, running analyses on BigQuery and producing visualizations to support investment decisions. • Focus Areas: Agentic AI, Generative AI-LLM, NLP, R&D, Machine Learning, Data Science • Stack: Python, SQL, GCP (BigQuery), Docker, Git, Redis (Pub/Sub, Streams), LangGraph, LangChain, OpenAI API, OpenAI Whisper, GPT-4o mini Transcribe, ElevenLabs, PaddleOCR, Vision-Language Models, FastAPI, Pydantic, asyncio.

  • Logiwa (Remote)
    • Senior Machine Learning Engineer
      Jul 2024 - May 2025 · 11 mos

      - Drove the AI roadmap for a cloud-native warehouse management platform, delivering features that helped logistics teams work smarter, from intelligent document search to predictive inventory insights. - Built a semantic search system using RAG & Generative AI that transformed how warehouse operators find information, replacing keyword searches with natural language queries. - Spearheaded open-source LLM adoption, deploying quantized models matching GPT-quality at a fraction of the cost • Focus Areas: R&D, Generative AI-LLM, NLP, Machine Learning, Data Science, ETL, Optimization • Stack: Python, SQL, Airflow, PySpark, Elasticsearch, Docker, PostgreSQL, Firestore, vLLM, Ollama, LangChain, LlamaIndex, HuggingFace, TensorFlow, PyTorch, FastAPI.

    • Machine Learning Engineer
      Oct 2022 - Jul 2024 · 1 yr 10 mos

  • Data Scientist at kariyer.net
    Dec 2020 - Oct 2022 · 1 yr 11 mos

    - Tackled a fascinating challenge at Turkey's largest job platform: matching millions of job seekers with the right opportunities through ML-powered systems. - Developed an intelligent resume parser using transformer models (BERT) that extracts structured information from messy PDF uploads. - Created content moderation models protecting platform integrity while reducing manual review workload. • Focus Areas: R&D, NLP, Machine Learning, Data Science • Stack: Python, SQL, Docker, TensorFlow, PyTorch, Transformers, Scikit-learn, FastAPI, Flask, New Relic.

  • Data Scientist at ICterra Information and Communication Technologies
    Jul 2019 - Dec 2020 · 1 yr 6 mos

    - Worked on cutting-edge 5G network security, developing anomaly detection models that identify suspicious patterns before they become security incidents. • Focus Areas: Network Security, Intrusion Prevention, Machine Learning, Data Science, Synthetic Data Generation • Stack: Python, TensorFlow, Scikit-learn, Pandas, NumPy.