Senior AI/ML Engineer/Data Scientist

Improved character identity consistency in multi-frame generations by fine-tuning LoRA models (kohya ss) and optimizing prompts; designed scalable workflow to automate training runs across GPU nodes. Built and deployed 20+ character-specific LoRA models using ComfyUI, cutting training cycles by 30% through curated dataset pipelines and minimal-step design. Owned and integrated a theme-based video generation API, enabling scalable, scriptable production of 100+ short videos used in downstream storytelling and animation workflows. Collaborated with product managers to define success metrics and iterate quickly.

Designed and delivered a scalable ML pipeline for resume parsing and ranking using transformers, LLMs, and classical models; reduced selection time from 3 days to 2 hours per 100 applicants. Took full ownership of architecture — including tokenizer logic, JSON output design, model deployment, and vector-based ranking system — ensuring modularity and reusability across HR teams. Led development of a production-grade chatbot to automate regional support, reducing manual requests by 40.

Spearheaded development of ML models for NFT price estimation, combining collection-based and global models to handle asset-level variability across major marketplaces. Achieved 90%+ T+1 prediction accuracy on millions of NFTs; optimized batch inference and latency for production API, ensuring real-time pricing capability. Mentored junior engineers and oversaw mid-level contributors, reviewing code, guiding experimentation, and enforcing best practices across modeling and data workflows.

Built and deployed NLP models for multilingual sentiment analysis, topic clustering, query understanding, and expansion; deployed features in AppGallery search at multi-million user scale. Collaborated with PMs and search infrastructure teams across 4 languages (Arabic, English, Spanish, Russian); designed modular APIs and lightweight pipelines for zero-downtime integration.

Preprocessed EEG data and developed ML models for cognitive signal analysis; supported instruction for graduate-level cognitive science courses.

Built and published a real-time Android messaging app (4.4/5.0, 1,000+ downloads) with custom emoji support and IRC-based backend.

Developed an image matching system using deep learning, reducing matching time by 91%; co-authored a related peer-reviewed paper.

Created an Android learning app for Turkcell Academy with animations, level-flow, and SQLite integration.