Comparative Analysis of Inference-Time Defense Methods for Multimodal Large Language Models
Bulat Nutfullin, Vladimir Evgrafov, Dmitry Namiot · arXiv preprint
CTO of the viju streaming service. Researching LLM-agent security as a PhD student at MSU CMC.
I am the CTO of Viasat Tech, leading a 35-person technology department (7 teams: Web, iOS, Android, Smart TV, Android TV, Go Backend, and systems analysis) that builds the viju streaming service for web, mobile, and TV platforms.
A CTO with technical roots: I started as a Go developer and grew into running an engineering department. I am good at quickly figuring out what is going wrong technically — in architecture, processes, or the team — and building a system that works. I have built an engineering department from scratch, migrated legacy systems to microservices, and I own the department budget, cutting infrastructure costs without sacrificing quality.
My profile sits at the intersection of engineering leadership, architecture, and AI. I manage development, DevOps, and the ML track, working with recommender systems, LLMs, and AI agents. In parallel, I am writing a PhD thesis at the Faculty of Computational Mathematics and Cybernetics of Lomonosov MSU and publishing research on LLM agent security and attacks on language models.
Open to part-time and project engagements as a fractional CTO or technical advisor.
Responsible for development, QA, infrastructure, architecture, and the AI track. In 2025: a hundred releases across 7 platforms, quarterly plan completion up from 40% to 83%, Android crash rate cut 14× (down to 0.17%), video start time halved, and a new product stream launched on all platforms simultaneously.
Second hire in the technology department: recruited 35 engineers, QA specialists, and analysts, and established the development processes. Designed and implemented a Go microservice backend — the service now sustains 100× the load at one quarter of the resource consumption.
Built a multi-service system for processing and interpreting medical data.
Developed a high-load, fault-tolerant ad serving system with a focus on safe handling of personal data.
Built microservices for communicating with various financial systems.
Worked on a monitoring service: agents collecting metrics from physical devices and individual processes on Linux.
Wrote and maintained the core Python codebase, built the software build and deployment system, and administered CI and monitoring.
Thesis on the security of LLM-based agentic systems.
Each paper has its own page with a short summary of what the work is about and why it matters. Full list on Google Scholar.
Bulat Nutfullin, Vladimir Evgrafov, Dmitry Namiot · arXiv preprint
VA Evgrafov, BM Nutfullin, DE Namiot · International Journal of Open Information Technologies 14 (5), 1-8
B Nutfullin, V Evgrafov · Tikhonov Readings, 121
VA Evgrafov, EA Ilyushin · International Journal of Open Information Technologies 9 (7), 21-31
Open to conversations about CTO / fractional CTO roles, architecture reviews, AI platform work, research collaboration, and focused technical consulting.