Burak Emre Tüylüoğlu

• Sole engineer on an academic author disambiguation (entity resolution) pipeline over bibliometric data, in production, architected toward 85M+ articles (~500M author mentions) on a single machine.
• Two-stage Python ETL: PostgreSQL 15 + pgvector (384-dim embeddings), Elasticsearch with a Turkish analyzer, SentenceTransformers, RapidFuzz.
• ~92% accuracy on a 200-case ground-truth benchmark, with cluster-quality guards.
• Cut full-pipeline runtime on a 2.5M-record dataset roughly in half (~4h → ~2h); largest win: O(N²) → O(N) keyset pagination.
• MinHash LSH block-then-cluster candidate layer that removed the Elasticsearch bottleneck; out-of-core memmap design keeping 85M-scale processing inside a fixed memory budget.
• Multi-signal matching hierarchy (email, ORCID, ResearcherID, affiliation, co-authorship graph, semantic embedding) with anti-snowball safeguards against false mega-clusters.

• Optimized YOLO-based detection models for inference speed and accuracy across several object detection problems: inference time down 28%, detection accuracy up 18%, real-time performance up 22%.

• Deep learning for computer vision: object detection, classification, pose estimation — fire-detection false alarms down 30%, plate-recognition accuracy 92% → 97%, pose estimation at 89% precision.
• Full cycle from data preparation and labeling through training and evaluation; served models via FastAPI backed by PostgreSQL, cutting retrieval time 60%.

• Unity game with AI-driven NPCs; integrated GPT so characters held unscripted conversations.
• Wired Azure text-to-speech and speech-to-text into a full voice interaction loop; speech recognition accuracy improved 88% → 94%.

• Internal automation tooling in Python; automated routine email workflows for HR and back-office teams, cutting manual workload by 75%.

Thesis: large-scale academic author disambiguation. First year done.

GPA 3.31. Senior project: an agentic assistant for Alzheimer’s patients — medication tracking and reminders, recording memories and photos and recalling them back in conversation, memory exercises.