Yagub Ismayilli

Evler.az LLC, IT Department · Baku, Azerbaijan

• Built an end-to-end handwritten text recognition (OCR) pipeline for Azerbaijani, a low-resource language, to automate digitisation of handwritten real-estate documents.
• Fine-tuned a Transformer encoder–decoder (TrOCR) on 25,000 synthetic Azerbaijani text-line images, reaching 6.40% character error rate and 55.4% exact-line-match accuracy.
• Collected and annotated a 5,000-image real handwriting dataset; a second fine-tuning stage cut character error rate to 3.47%, word error rate to 10.4%, and raised exact-match accuracy to 75.7%.
• Benchmarked against TrOCR-small (21.44% CER) and a CNN–BiLSTM–CTC baseline (17.23% CER) — roughly a fivefold error reduction.

GNI Software · Baku, Azerbaijan

• Fine-tuned DialoGPT-small, a 117M-parameter GPT-2 variant, into a customer-support chatbot using Hugging Face Transformers and the Trainer API.
• Built the training corpus from the 3M-tweet Twitter Customer Support dataset, reconstructing conversation threads to extract roughly 20,000 single-turn prompt–response pairs.
• Trained within 4 GB of VRAM (RTX 3050) using mixed-precision (fp16) and small-batch training; loss converged from 4.6 to 3.3 with no signs of overfitting.
• Profiled the corpus by token length, intent category and vocabulary, and documented limitations around single-turn context, single-brand domain bias, and lack of factual grounding.

Freelance · Remote

• Developed and fine-tuned NLP models for text classification, sentiment analysis and sequence processing using TensorFlow / Keras.
• Trained and evaluated supervised machine learning models for prediction and classification tasks, covering feature engineering, model selection and validation.
• Cleaned, transformed and analysed client datasets in Python (Pandas, NumPy), and built responsive web applications with React.js and Node.js.

Meridian Course · Baku, Azerbaijan

• Taught core programming concepts, data structures and algorithms through real-world examples and structured lesson plans.
• Reviewed student code and guided debugging, building their technical and analytical problem-solving skills.

Lider Advertising · Baku, Azerbaijan

• Developed and maintained internal web applications using Laravel and PHP, backed by a relational database, and wrote SQL queries for diagnostics and reporting.
• Provided technical support to internal users, resolving production system issues and improving day-to-day workflow efficiency.
• Debugged and optimised existing application code, improving reliability and reducing recurring system failures.

Azerbaijan-French University (UFAZ)

Thesis: Development of a Handwritten Text Recognition (OCR) System for the Azerbaijani Language

University of Strasbourg

Thesis: Development of a Handwritten Text Recognition (OCR) System for the Azerbaijani Language

Khazar University

82.57 / 100