LUMA

Context

LUMA is my own B2B AI-solutions agency, based in Russia. We were among the first in the region to treat AI not as a gimmick but as operational infrastructure — building systems that automate genuinely complex, cross-functional processes across finance, marketing and business development. The thesis is simple: most companies run critical workflows by hand across a dozen disconnected tools; LUMA turns those into autonomous, AI-driven pipelines that run themselves.

How we work

Every engagement follows the same arc — map the manual process end to end, find where human time is burned on low-value repetition, then design a pipeline that ingests the data, applies LLM reasoning where judgment is actually needed, orchestrates the steps, and delivers a finished output — leaving a human only at the decision points that matter.

Selected builds

• Elizium — an autonomous content factory for a major regional TV network. A full AI pipeline that produces broadcast content at scale: it takes topics and briefs at the input, generates and formats the content with LLMs, adapts it to the network's formats, and schedules it into the publishing calendar automatically. What used to need a standing content team became a largely self-running line.
• Telegram → 1C management reporting — employees submit performance metrics through a Telegram bot in seconds; the bot structures and validates the input and pushes it into 1C, which auto-compiles reports for leadership. Replaced manual data-chasing and spreadsheet assembly with a near-real-time reporting loop.
• AI lead-qualification assistant — an LLM-driven first-touch layer that reads inbound leads, qualifies and tags them, and routes each to the right owner.
• Automated KPI dashboard — pulls data from CRM and ad networks into a single live management view, replacing scattered manual reports with one always-current source of truth.

Impact

Across builds the pattern repeated: multi-day, multi-person reporting and content cycles collapsed to minutes, manual operations were largely engineered out, and clients got first-mover automation in a market where almost no one was doing this yet.