Resume

Resume

Emirhan YILDIRIM

Applied ML EngineerComputer Vision · LLM Systems · Production MLOps

Summary

Information Systems Engineering graduate specializing in Applied ML, Computer Vision, and production MLOps. Experienced in transitioning deep learning models from prototype notebooks to containerized, high-throughput production environments. Shipped medical AI classifiers with inference-time domain adaptation, RAG-grounded assistants, and hybrid recommendation engines. Seeking remote-first Applied ML / AI Engineer roles where engineering robust, high-performance systems is the core focus.

Experience

Software Engineer Intern | ESBI Bilisim

Jul 2026 - Present | Sakarya, Türkiye

  • Collaborating on backend architecture, focusing on the design and optimization of enterprise APIs.
  • Engineering robust data processing logic and database schemas to handle transaction workflows.
  • Implementing structured system logging and fault-tolerant mechanisms for stable service operations.

Software Engineer & Founder | Zidorun (Personal Venture)

Jun 2025 - Present | Sakarya, Türkiye

  • Founded and engineered the backend infrastructure for a high-volume digital collectibles platform operating across multiple regional C2C marketplaces.
  • Designed a modular, config-driven inventory management and distribution service in Python, decoupling platform-specific logic from core database states using JSON schemas.
  • Engineered fault-tolerant data pipelines with integrated telemetry and Telegram alerts, reducing synchronization latency by 90% and successfully processing 700+ transactions with a 95%+ success rate.

Software Engineer Intern | Platech Coating

Jun 2025 - Jul 2025 | Sakarya, Türkiye

  • Developed a modular desktop application in Python to automate complex operational and ERP data synchronization workflows.
  • Designed a multi-threaded system architecture with fault-tolerant retry loops and logging, ensuring stable data extraction and parsing of XML, XLSX, and TXT feeds.
  • Built a local mock ERP server environment (test harness) to simulate server-side network dropouts and exception states for offline validation.
  • Impact: Reduced daily data processing cycle times by 90% (saving 4–6 hours of manual entry down to under 30 minutes); system remained in active production use post-internship.

Projects

WBC Analyzer: AI-Powered Pathology Assistant

Jan 2026 - Apr 2026

  • Engineered an end-to-end WBC classification system using DenseNet121, a custom 5-step Medical Enhanced Filter preprocessing pipeline, WBCAttention, and MedSwish activation, achieving 98.53% in-distribution accuracy on a 7.83M parameter model.
  • Demonstrated robust OOD generalization with 89.05% accuracy (+32.09 pp boost) on cross-lab peripheral blood smears via inference-time domain adaptation (Binary Routing + Reinhard normalization + TTA) without retraining.
  • Built and containerized a production-ready Flask REST API with multi-stage Docker build, GitHub Actions CI pipeline (automated test + image build on push), structured request metrics via /health endpoint, and fault-tolerant error handling delivering real-time responses in 200–350 ms.
  • Integrated a separate Grad-CAM observer agent (GPT-4o & Gemini 2.5 Flash) that reviews visualization outputs against scene-specific prompts to flag shortcut learning, decoupled from the core inference pipeline.
  • Published preprint on ResearchGate: doi.org/10.13140/RG.2.2.34201.79208

Kinematic Action Recognition: Industry 5.0 Pipeline

Apr 2026 - May 2026

  • Built an end-to-end ML pipeline on 10 GB motion-capture data (132 sensor channels) utilizing out-of-core ingestion with Dask and sliding-window feature extraction yielding 528 features per window.
  • Implemented three model tiers: unsupervised clustering (MiniBatchKMeans + PCA + Hungarian matching), supervised classification (RandomForest, Macro F1: 0.9995), and real-time streaming (ARF + ADWIN drift detection at 81 windows/sec, 59 MB peak RAM).
  • Identified and resolved temporal leakage in overlapping windows by replacing random split with a systematic every-5th-window split.
  • Competition variant: engineered kinematic features (elbow joint angles, bilateral joint distances, rotation norms) and built a LightGBM + RandomForest soft-voting ensemble, achieving 0.94169 accuracy on private leaderboard.

RAG Project Assistant: Hybrid-Source Q&A System

Jun 2026 - Present

  • Built a Retrieval-Augmented Generation system answering questions about five personal ML/automation projects, combining AST-extracted code structure with hand-curated documentation in a hybrid knowledge base.
  • Engineered a similarity-threshold gate (ChromaDB, L2 distance < 1.40, empirically calibrated) paired with a strict grounding prompt requiring explicit source attribution to suppress hallucination on out-of-scope queries.
  • Resolved cross-project retrieval contamination via keyword-based project detection and metadata filtering, with full-corpus fallback when filtered retrieval returns no results.
  • Deployed as a FastAPI service on Hugging Face Spaces (Docker), migrating inference from local Ollama to Groq’s Llama 3.3 70B; added IP-based rate limiting (slowapi) and a public JS chat widget embedded on portfolio site.

Popcorn Wagon: Hybrid Movie Recommender System

Jan 2025 - May 2025

  • Built a full-stack Flask recommendation platform combining content-based filtering (TMDB API) with collaborative filtering (MovieLens + SVD + Annoy).
  • Engineered millisecond-level similarity search using Spotify’s Annoy for nearest-neighbor lookups over the trained SVD embedding space.
  • Built scalable ETL pipelines with Pandas and Dask for dataset cleaning and preprocessing.
  • Designed normalized SQLAlchemy schema with user authentication, session management, and personalized list/tagging features.

Tech Stack

Programming

Python, SQL

AI / ML

PyTorch, TensorFlow, Keras, Scikit-Learn, OpenCV, XGBoost, LightGBM

LLM & RAG

ChromaDB, RAG Pipelines, Sentence Transformers, LLM APIs (Groq, Ollama)

MLOps & Systems

Docker, GitHub Actions, CI/CD, CUDA, Git, Model Observability

Backend & Data

FastAPI, Flask, Dask, Pandas, REST APIs, NumPy

Education

BSc in Information Systems Engineering
Sakarya University | Sep 2021 - Aug 2026

Languages

  • Turkish: Native
  • English: Professional

Certificates

  • CS50x: Introduction to Computer Science - Harvard University
  • Supervised Machine Learning: Regression and Classification - DeepLearning.AI