AI engineer — forward-deployed agent systems
Artem BorisovI build and operate production AI agent systems solo — discovery, architecture, code, deployment and day-two operations.
Running in production in
Selected work
Systems now running in production
Clients are described by industry rather than by name — the numbers below are their operating data, and it is not mine to publish. Each case links to what the problem was, what I built, and what changed.
A knowledge base that answers agronomists from current research
24/7access to a current knowledge base
Project documentation from a text description, straight to the cutter
×10faster documentation
How I work
Four rules I got the expensive way
Find the bottleneck before writing code
Twenty years of operating businesses and 300+ in-depth customer interviews are why I start in the field, not in an editor. Sometimes the answer is that the process itself is broken — at the surf school I lifted 15-minute response coverage from 60% to 79% with people first, and only then built the agent on top of a process that worked.
Judgement to the agent, arithmetic to code
In the estimating system the model sits at exactly one point — reading unstructured input: a photo of a sketch, a voice note, a PDF. Everything touching money is deterministic code. In a task that has to be right to the last unit, an LLM doing the arithmetic would be both more expensive and worse.
A human stays on the money
Guardrails on every model call, a budget ceiling, output validation. When input is thin the agent asks a question instead of inventing an answer, and anything that moves money or touches a customer is confirmed by a person. Failures surface immediately rather than quietly — silence from a monitor is not health.
I run what I ship, then hand it over
Deployment, systemd, monitoring and day-two operations are part of the job, not someone else's phase two. And the goal is to become unnecessary: at the language school I trained the finance manager to extend the system herself, and it now evolves without me.
Technical skills
What I actually work with
AI / agents
- Python
- Multi-agent systems (Google ADK, Claude Agent SDK)
- A2A, MCP
- RAG
- Tool orchestration and state management
- Guardrails, evals
- Prompt and context engineering
Integrations
- amoCRM
- 1C ERP
- Google Workspace API
- Telegram Bot API
- Instagram Graph API
- Speech-to-text pipelines
Infrastructure
- Linux, systemd, Caddy, nginx
- git / GitHub
- Production deployment
- Day-two operations
Discovery
- JTBD interviews
- Customer development
- Unit economics
About
Twenty years of running businesses, then two of building for them
Over the past two years I have shipped AI and automation systems now running in production for clients in tourism, electrical contracting, education, legal services and seafood processing.
Before that: 20 years operating businesses as a director and co-founder, and 300+ in-depth customer interviews. That is why I find the real bottleneck before writing code — I have sat on the other side of the table, paying for software that solved the wrong problem.
Earlier career
2022 — present
Independent consultant
Helping business owners structure their processes and adopt AI. Since 2024 this has meant building and running the agent systems myself.
2018 — 2022
Marketing Director and co-founder, agricultural biotech company
Prepared the due diligence package for an investment round; adapted the product for the US market and took part in the raise. Launched a new retail product line after 20 farmer interviews. Still a co-founder.
2008 — 2022
Commercial Director, private language school
Grew it from a two-room flat into one of the largest private schools in the region. Commissioned and owned its CRM, its integrations and a loyalty programme that cut discounts from 7% to 2%.
2005 — 2020
Director, cable TV and internet holding
Ran four TV channels, a newspaper and an ad agency; team of 15. Commissioned and owned custom broadcast software that stayed in production for 11 years; launched a self-service classifieds product with payment and automatic hand-off to the broadcast pipeline.
Teaching and speaking
Workshops on AI agents for entrepreneurs
Hands-on sessions on putting agents into real business processes; JTBD masterminds and accelerator webinars.
Guest lecturer, HSE University
Art and Design School — product research lectures on the «Fashion Industry. Product Management» programme.
Expert, Be In Open institute
Webinars on product research and applied AI for fashion brands and founders.
Education
Kamchatka State Technical University, 2006
Specialist degree (5-year) in Applied Informatics, with honours.
Continuing
GoPractice (data-driven product management) · «How to Build Products» (Zamesin) · Tallinn School of Managers, with honours.
Languages and community
Russian — native. English — B1 (Cambridge exam): confident reading and listening, functional spoken English, actively improving.
Regional lead of the international Let’s Do It clean-up movement in 2018 — brought 6,000+ volunteers onto the streets in one day; commendation from the regional governor.
Based in Russia (UTC+12), open to relocation.
Let’s talkabout your bottleneck
Hiring, a project, or a second opinion on an agent design — Telegram is the fastest way to reach me. I read everything myself.
mail [at] artemborisov.ru
Based in Russia (UTC+12) · open to relocation