I've built production NLP, RAG, and LLM-agent systems across Neuro.net, GRONICS, and Fortech, owning delivery from data and evaluation design through deployment and observability.
At Neuro.net, I designed and deployed an agentic assistant for internal support teams using query classification, hybrid RAG, tool calling, human approval, and stateful CRM and Service Desk workflows. I established offline evaluation and automated approximately 35% of standard incoming requests.
At GRONICS, I built an end-to-end RAG assistant for engineers and support specialists, improving Recall@10 from 0.68 to 0.86 and MRR@10 from 0.54 to 0.74. I also reduced median retrieval-plus-generation latency from approximately 3.8 seconds to 1.7 seconds.
At Fortech, I trained Transformer models for classification and named-entity recognition, improving macro-F1 from 0.79 to 0.89. I optimized production inference from 260 ms to 78 ms CPU p95 latency and reduced manual request routing by 42%.
