Albert Hakobyan

I am an AI Engineer specializing in Natural Language Processing, low-resource language modeling, enterprise Retrieval-Augmented Generation (RAG), and autonomous agentic workflows. Holding a B.S. in Data Science from the American University of Armenia and DataCamp's Associate AI Engineer certification, I bridge theoretical machine learning with production-ready software engineering. My research on neural punctuation systems for Armenian won the AUA ACSE Research Poster Showcase and was accepted at CODASSCA 2026.

I design end-to-end Python systems, focusing on RAG engines that operate over massive knowledge bases (100GB+) with exact source citations, soft scope routing, and zero-hallucination auditing. I architect multi-agent systems using LangGraph and LlamaIndex, as well as custom n8n automations for marketing, lead enrichment, and complex data workflows. My projects feature live web discovery, automated media transcription, and human-in-the-loop control gates.

Beyond orchestration and automation, I specialize in fine-tuning (ML/DL/NLP), vocabulary adaptation, and knowledge distillation. By distilling massive models like Gemini 2.5 Flash into compact mBERT and BiLSTM ensemble learning solution or performing LoRA surgery on Qwen2.5, I build lightweight local systems running at 1,000+ sentences per second on CPU at zero inference cost.

Whether you need complex agentic automations, marketing or other workflow integration via n8n, high-precision semantic search, or model optimization to slash cloud infrastructure bills, I deliver robust, Dockerized FastAPI microservices ready for seamless deployment.

I am proficient in English, Russian and Armenian languages.