Senior ML Engineer

На данный момент Senior ML Engineer.

Мои компетенции и опыт

Senior AI/ML Engineer with 15 years of experience (8 in ML), turning cutting-edge research into production systems- from Edge Computer Vision on 30M+ devices to modern LLM architectures, RAG and Agentic AI. Hands-on across the full stack, from model training and fine-tuning to low-level optimization and cloud deployment. Led a cross-functional team of 7 for 6+ years.

KEY PROJECTS AND ACCOMPLISHMENTS:

• Fine-tuned SOTA LLMs (LLaMA, Gemma) using Hugging Face Transformers and PEFT (LoRA/QLoRA) for human preference modeling (reward modeling in RLHF), achieving up to 15% improvement in Log Loss over zero-shot baseline.
• Built a RAG system for legal QA over 300+ documents. Implemented agentic ingestion with Docling, VLMs and Qdrant; hybrid retrieval using embeddings, BM25 and semantic re-ranking; and cross-document reasoning with LangGraph. Deployed self-hosted inference with Qwen3.5 via vLLM, achieving нужен доступ к резюме TTFT.
• Led an R&Dproject to develop an image enhancement neural network (U-Net, PyTorch), building an end-to-end ML pipeline from dataset collection to on-device deployment; achieved <800ms inference for 12MP images on smartphones via quantization and porting model to Qualcomm Hexagon DSP for Edge AI deployment.
• Led the development, integration, and support of a best-in-class Super Resolution SDK (C/C++), deployed on 30M+ devices in 50+ smartphone models. Developed the world’s first real-time video SR SDK for mobile platforms, running at up to 60 FPS. Several smartphone models powered by the SDK achieved the top 1 DxOMark ranking upon release.
• Contributed to open-source LLM tooling: added Gemma2 support to AutoGPTQ, improved LangChain-Google integration (bug fixes, refactoring).