Discuss a project

Dmitry Komarov

Independent AI/LLM architect · Former Head of AI in banking

Secure Enterprise AI. From pilots to production.

I help companies turn GenAI ideas into measurable production systems: LLM platforms, agentic workflows, AI security, automation, observability, cost control, and governance.

Message on TelegramFill the briefSee services
Dmitry Komarov
Head of AI in Rosbank; AI Lead across the T-Bank transition
Built AI platforms and production LLM workflows for regulated environments
Managed AI/product analytics teams of 30+ specialists
Python since 2013; TypeScript, infrastructure, local LLMs, servers and AI hardware
Published on LLM injection and AI security at Hightech.fm
AI Engineer at The Value Engineering: LLMs, agents, automation, AI readiness

services

How I can help

I can help at different levels of intent: a quick expert review, a readiness audit, an LLM/agent security review, production development, or ongoing AI leadership.

Fast entryAI/LLM Consultation60-90 min · Fast startA focused review of your AI task, architecture, vendor choice, workflow, or security risk.
  • Clear next step
  • Architecture and risk notes
  • Audit or build recommendation
Primary offerAI Readiness & ROI Audit300-500k RUB · 1-2 weeksA practical audit of processes, data, use cases, risks, and ROI potential before you spend months on another AI pilot.
  • Use-case map
  • Risk and data readiness score
  • Roadmap with business impact
DifferentiatorLLM / Agent Security AuditBy scope · 2-3 weeksPrompt injection, tool misuse, data leakage, MCP/A2A risks, guardrails, access control, and auditability.
  • Threat model
  • Attack surface review
  • Guardrails and observability plan
BuildAgent / Automation DevelopmentPilot to production · 4-10 weeksOne workflow taken from problem framing to integrated AI system with tools, evals, monitoring, and human handoff.
  • Working pilot
  • Production architecture
  • Evals, tracing, and cost controls
StabilityFractional Head of AIMonthly retainer · 3-6 monthsAI leadership for companies that need senior judgment before hiring a full internal AI function.
  • AI roadmap
  • Vendor and architecture decisions
  • Team enablement and governance

writing

Writing and field notes

I write about the parts of AI work that decide whether a system survives production: context, evals, tool risk, security, local models, and operating discipline.

2026-05-23 · 4 min readWhy enterprise AI breaks between demo and productionEnterprise AI usually breaks after the demo, when the system meets real data, permissions, workflows, evals, security, cost, and ownership.2026-05-19 · 2 min readAI Readiness Audit: how to find where LLMs and agents can actually create ROIBefore building another AI pilot, map the process, data, risk, operating model, and ROI path. This is the audit I run before recommending a build.2026-05-19 · 2 min readContext engineering for enterprise agents: why prompts are not enoughEnterprise agents fail when teams treat context as a long prompt. Production systems need scoped memory, retrieval, permissions, tool context, evals, and observability.
Read all articles

process

How we start

  1. You send a short brief or Telegram message.
  2. I clarify the business goal, risks, data context, and whether I can help.
  3. If there is fit, you get a concrete next step: consultation, audit, security review, pilot, or longer-term AI leadership.