Summary
Nikita Parfenov is an AI & ML Solution Architect with 8 years of experience and 5+ years focused on building production-grade ML and LLM systems, specializing in semantic matching, ranking, information retrieval and RAG architectures. He has designed low-latency, scalable ML pipelines and inference stacks for high-throughput environments, including programmatic advertising systems handling 10–15 TB/day and industrial forecasting models that supported multimillion-dollar decisions. Nikita has repeatedly moved LLM projects from prototype to production—optimizing inference with vLLM/Triton, fine-tuning models for noisy real-world inputs, and delivering 96–98% entity-matching accuracy in FMCG receipt/barcode tasks. Comfortable across cloud MLOps (AWS, Azure), PyTorch/HuggingFace, and vector retrieval frameworks, he blends hands-on engineering with client-facing architecture and team onboarding. A former reservoir analytics lead, he brings uncommon domain depth in uncertainty-aware forecasting and decision analytics that informs his pragmatic approach to ML robustness. Based in Tbilisi, he balances practical production delivery with playful humility—“programming on ctrl+c ctrl+v”—and a knack for turning messy operational data into reliable AI services.
8 years of coding experience
3 years of employment as a software developer
Master of Science (BSc + MSc), Physics, Master of Science (BSc + MSc), Physics at Tomsk State University
Heriot-Watt University Edinburgh Campus
Artificial Intelligence (professional retraining diploma), Data Scientist, Artificial Intelligence (professional retraining diploma), Data Scientist at GeekBrains (e-learning)
Physics, Physics at Institute of Strength Physics and Materials Science SB RAS