Skip to content
Kerem Kurban

Geneva, Switzerland

Kerem Kurban

ML Engineer | Agentic AI, Evaluation & Interpretability

AI/ML engineer and computational neuroscientist. I build and evaluate agentic AI systems in production, including in regulated industry, and study how models arrive at their outputs, from brain circuit models to multimodal medical foundation models.

Agentic SystemsLLM EvaluationInterpretabilityMedical Foundation ModelsComputational NeurosciencePython
Download CVSee projects
Profile photo of Kerem Kurban

01 · Experience

Experience

  1. Machine Learning Engineer (Forward Deployed)

    11/2025 – 06/2026

    Wipro, forward deployed to a large Swiss bank

    • Embedded on-site within the client's Core Engineering Data & AI function (~60% client travel), managing concurrent engagements across twelve reconciliation domain groups plus risk and operations stakeholders; led discovery, solution framing, and iterative delivery from proof-of-concept to production. Initial six-month contract extended in recognition of contributions
    • Built a multi-agent benchmarking harness evaluating autonomous agents against historical human reconciliation decisions, integrating MLflow 3 for experiment tracking and observability, Unity Catalog for governance, and a purpose-built MCP server for governed data retrieval; established audit-ready AI quality assurance in a regulated financial environment
    • Designed and shipped an agentic ingestion pipeline (Azure Document Intelligence + vision-language models, Pydantic schema-guided extraction, parallel agentic execution with cross-validation of generated artefacts) standardizing 330 multimodal SOPs into a governed data asset, cutting manual review effort by ~90%
    • Extended the client platform with custom tooling: a React application exposing SOP coverage and automatability scores as a self-serve data product, and a domain ontology / knowledge graph linking reconciliation break types to SOPs for provenance-aware, auditable agent reasoning
    • Led a bank-wide evaluation of enterprise AI/ML platforms and model registries; proposed the solution architecture that shaped programme platform strategy; ran cross-functional enablement sessions for business and technical stakeholders, from engineers to senior executives, and mentored a junior engineer

  2. AI Researcher (Contract)

    03/2025 – 10/2025

    Neptune.ai

    • Post-training and alignment on Mistral, Qwen3 and Llama 3.1 8B: supervised fine-tuning, DPO, PPO, GRPO
    • Benchmarked PEFT methods (LoRA, Q-LoRA, DoRA), reaching a 210% improvement over baseline

  3. Machine Learning Engineer

    04/2024 – 01/2025

    EPFL Blue Brain Project, Geneva

    • Built an LLM evaluation pipeline (RAGAS plus custom tool-calling benchmarks) measuring hallucination rates across retrieval configurations
    • Engineered a modular Python toolkit (Pydantic, LangChain, async APIs) for cross-modal extraction and knowledge-graph search over heterogeneous scientific data, cutting manual research effort by roughly 90%
    • Built a Neo4j knowledge graph over BBP citation networks with GraphRAG retrieval and a domain-expert-facing assistant
    • Co-developed Neuroagent for in-silico simulation and hypothesis testing of biophysical brain models on AWS, now available at openbraininstitute.org

  4. Graduate Researcher & Data Scientist

    04/2021 – 04/2024 (intern 03/2020 – 09/2020)

    EPFL Blue Brain Project, Geneva

    • Dedicated engineer for network building, atlas alignment and in silico wet-lab replication on the full-scale CA1 model (PLoS Biology 2024)
    • Deployed Dask-parallelised network statistics on large-scale graphs, and Spark pipelines over terabyte-scale simulation output on CSCS HPC under Slurm

  5. Graduate Research Assistant

    09/2018 – 03/2021

    UNAM, Bilkent University, Ankara

    • Biologically realistic spiking network models (Izhikevich, LIF, MAT with STDP) in NEURON and NEST
    • Evolutionary optimisation of SNN controllers for locomotion pattern generation

  6. Machine Learning Ambassador

    05/2017 – 09/2018

    Intel, Istanbul

    • Designed machine learning models for analyzing fMRI data
    • Showcased neuroscience research on Intel's AI platform

  7. Software Engineer Intern

    06/2016 – 09/2016

    Neurolize

    • Performed marketing research using EEG and eye tracking
    • Analyzed online shopping interactions using classification techniques

02 · Projects

Featured Projects

Showing 8 projects

  • Biophysically Detailed Model of Rat Hippocampus CA1 Region

    Developed and maintained in-silico models of detailed neurons in 3D rat atlas, validated by in-vivo and in-vitro experiments and provides insights into hippocampal function from its structure, physiology and connectivity.

    PythonneuroscienceHPCgraph-theory
    GitHub repository for Biophysically Detailed Model of Rat Hippocampus CA1 Region

  • Sonata to Neo4j

    Convert Biophysical Neuron Models and Simulations into Neo4j Graph and create fact sheets.

    PythonNeo4jneuroscience
    GitHub repository for Sonata to Neo4j

  • Sortify

    Sortify is a web application that allows users to sort their favorite Spotify album tracks based on their preferences.

    PythonHTML/CSS
    GitHub repository for Sortify

  • Scholarag

    A Retrieval Augmented Generation (RAG) API meant for scientific literature, which includes data management utilities and relevant endpoints for efficiently showcasing papers to your users.

    RAGPythonAWSDocker+1
    GitHub repository for Scholarag

  • Hypnosis Simulator

    An interactive web application featuring various hypnotic visualizations and audio stimulations for relaxation and meditation. Includes multiple visualization types like Spiral Induction, Pulsing Light, and 3D Lorenz patterns, with customizable controls and audio accompaniment.

    JavaScriptThree.jsWebGLInteractive+3
    GitHub repository for Hypnosis Simulator Live demo of Hypnosis Simulator

03 · Publications

Publications

Download all publications
  1. Poster2018

    Machine and deep learning approaches for the study "Effects of early anesthesia exposure on human brain development using multimodal neuroimaging"

    Kurban K., Budur E.

    16th National Neuroscience Congress in Turkey

    machine-learningfMRIneuroscience
  2. Poster2024

    Characterizing subtypes of projecting axons in mice using topological data analysis and machine learning

    Kurban K., Kanari L.

    Society for Neuroscience, San Diego, CA, USA

    machine-learningneuroscience
  3. Journal2024

    A connectome manipulation framework for the systematic and reproducible study of structure–function relationships through simulations

    Pokorny, C., Awile, O., Isbister, J. B., Kurban, K., Wolf, M., & Reimann, M. W.

    bioRxiv

    simulationneuroscience
    Read paper: A connectome manipulation framework for the systematic and reproducible study of structure–function relationships through simulations
  4. Poster2023

    A deep dive into CA1 network: Insights from Network Science

    Kurban K, Romani A., Markram H.

    32th Computational Neuroscience Society

    hippocampusneurosciencegraph-heory
  5. Preprint2023

    Community-based Reconstruction and Simulation of a Full-scale Model of Region CA1 of Rat Hippocampus

    Romani, A., et al.

    Cold Spring Harbor Laboratory

    simulationneurosciencehippocampus
    Read paper: Community-based Reconstruction and Simulation of a Full-scale Model of Region CA1 of Rat Hippocampus
  6. Poster2022

    Topological properties of full-scale model of rat hippocampus CA1 and their functional implications

    Kurban K, Pokorny C., Romani A.

    Society for Neuroscience, San Diego, CA, USA

    hippocampusneurosciencegraph-heory
  7. Journal2020

    Resting-state network dysconnectivity in ADHD: A system-neuroscience-based meta-analysis

    Sutcubasi B, Metin B, Kurban MK, Metin ZE, Beser B, Sonuga-Barke E

    World J Biol Psychiatry

    meta-analysisneurosciencefMRIADHD
    Read paper: Resting-state network dysconnectivity in ADHD: A system-neuroscience-based meta-analysis

04 · Certifications

Certifications

  • Neo4j Certified Professional

    Neo4j GraphAcademy

    2024

  • LLMOps

    Deeplearning.ai

    2024

  • AI Agents in LangGraph

    Deeplearning.ai

    2024

  • Fine Tuning Large Language Models

    Deeplearning.ai

    2024

  • Quantization Fundamentals in Hugging Face

    Deeplearning.ai

    2024

  • AWS Foundations

    Amazon Web Services

    2024

  • AWS Bedrock

    Amazon Web Services

    2024

  • Infrastructure as Code in Google Cloud Platform

    LinkedIn Learning

    2024

  • LLMs as Operating Systems: Agent Memory

    Deeplearning.ai

    May 2025

  • Federated Fine-Tuning of LLMs with Private Data

    Deeplearning.ai

    May 2025

  • Evaluating AI Agents

    Deeplearning.ai

    May 2025

05 · Contact

Get in Touch

keremkurban@proton.meLinkedIn ProfileGitHub ProfileDownload CV

Geneva, Switzerland