Enes Karanfil

Senior Machine Learning Engineer

n11•  January 2024 - Present

Incorporate Retrieval-Augmented Generation (RAG) for a more comprehensive understanding of product titles, descriptions beyond exact keyword matches. Designed and implemented a visual similarity search system with under 1% false positive rate, enhancing fashion product recommendations. Utilize LangChain and LLM (LLaMA 3.1) to extract and evaluate relevant details from product reviews in the e-commerce domain, providing abstract summaries so users don’t have to read over 1000 reviews for a single product. Developed and deployed a high-performance system using microservices for embedding generation, indexing, and nearest neighbor search by using vector database. Conducted comprehensive analysis of business requirements and experimented with solutions from recent publications. Developed and integrated a spell checker model, reducing zero-result page metrics by approximately 1.5% across Web, Android, and iOS platforms. Ensured the system’s high-load handling capabilities and optimized performance, achieving significant improvements in recommendation accuracy.

Machine Learning Engineer/Applied Scientist

Trueyogi•  March 2022 - January 2024

Played a key role in decision-making at Trueyogi, a small-scale startup, collaborating closely with backend and iOS teams, and providing high-load ML endpoints. Additionally, led and mentored junior ML engineers and interns. Developed a human pose checker module using a lightweight machine learning model integrated on edge devices, achieving an AUC score of ≈ 0.8. Created a human face analysis module capable of detecting lesions, acne, and wrinkle locations using unsupervised learning techniques without requiring labeled data, reaching ≈ 5000 subscribed users.

Machine Learning Engineer

Huawei•  April 2021 - March 2022

Collaborated closely with AI researchers to implement lemmatization for Arabic and Turkish languages within Elasticsearch services, improving search accuracy and relevance. Created and optimized efficient Elasticsearch queries to enhance data retrieval and search performance. Implemented distributed training on Huawei Cloud, optimizing the performance and scalability of machine learning models. Worked on a content moderation project by fine-tuning DistilBERT for specific downstream tasks, achieving over 95% accuracy.

Machine Learning Engineer

GarantiBBVA Technology•  March 2020 - April 2021

Designed and implemented a machine learning prediction pipeline capable of managing high-throughput data streams. Processed over 300,000 data points per minute effectively, ensuring system reliability and scalability. Developed efficient data processing pipelines for anomaly detection models. Created applications in Scala and Spark to handle data processing and analysis tasks.

Hacettepe University

Master of Science in Computer Science•  January 2024 - January 2024•  CGPA: 3.83

My research focuses on integrating interleaved optical and spectral remote sensing image data into the LLaVA framework, aligning these multimodal inputs with text embeddings during the pre-training phase. Additionally, fine-tuning a decoder-only large language model using a curated dataset, which represents a significant contribution to the training process.

TOBB University of Economics and Technology

Computer Science, Bachelor of Science in Computer Engineering•  January 2015 - January 2021•  GPA: 2.81

Roadpulse is a real-time application designed to detect road potholes using camera streams.