Jyri Gromyko LintunenGG Solutions Oy
Subcontracting · LLM and AI

AI features from demo to production

I run several LLM pipelines in production: email classification and draft answers, sourced company dossiers, computer vision for job costing and receipt processing. Prompts live versioned in a registry where they can be run as evals, and the cost of calls is logged. I bring the same discipline to your project.

  • Within a weekstart
  • €80/h excl. VATrate
  • Claude, Gemini, OpenAImodels
  • Remote or hybridmode
Tell me about the AI task Download CV (PDF) WhatsApp

Projects that relate to this

Three systems where a language model or computer vision does real work every day.

B2B CRM & lead automation

Architect & sole developer · In production · 2026 to present

A sales system where AI handles the mail: every campaign reply is labelled and summarised, and a draft answer is prepared in the same thread.

  • Inbound triage: rules drop the noise, a language model judges the rest, and leads are registered in the system without manual work.
  • Company dossiers with Claude Opus: every claim points to its source by id, and the report comes out as a finished PDF.
PythonFlask 3.1SQLAlchemy 2.0PostgreSQL 16React 19Vite 8CeleryRedis
Why it matters: Several LLM pipelines running side by side in production, not demos.

SKT quoting software: computer vision for job costing

Architect & developer · Active client project · 2026 to present

Building assessment and window-cleaning cost estimation for the Helsinki metro area

  • A 10-phase pipeline: Oikotie scraper (cards, details and photos) → Digitransit geocoding → Google Street View (URL-signed) and Mapillary → Claude Sonnet Vision → man-hour formula v2.0 with a v3.0 refinement → a proposal from a single command.
  • Cost tracking for every LLM call with its own self-test, so the price of vision analysis stays visible before a run is scaled to hundreds of targets.
10pipeline phases1,7 GBof imagery to analyse
Python 3.11Claude Sonnet VisionGoogle Street ViewMapillaryDigitransitSQLiteStreamlitDocker
Why it matters: Computer vision inside a pipeline, with the cost known before a run is scaled.

Personal Advisory Council

Architect & developer · Working prototype · 2026

A self-hosted multi-AI advisory board that evaluates decisions in parallel

  • A five-stage workflow: ingest → an architect agent proposes the council → interactive editing → parallel execution → peer review and synthesis.
  • Real-time token and cost tracking, configurable reasoning depth and live progress tracking during a run.
  • A Docker Compose stack: FastAPI back end, React front end, a LiteLLM proxy, ChromaDB for vector search and an optional Telegram bot.
8agent personas in parallel
Python 3.12FastAPISQLAlchemy 2.0openai-agents SDKLiteLLMChromaDBReact 19Tailwind 4
Why it matters: Agents in parallel, LiteLLM, ChromaDB and live cost tracking.

How we start

In AI work the first week is often measurement: test cases, today's accuracy and the price per call, before anything is changed.

Send the task and the stack

In a few lines: the project, the technologies, the timeline. I answer the same working day with my CV and the two closest examples of my work.

A scoped first week

I start with a small, clearly bounded task in your repository and your tools. At the end of the week you see the code and decide what happens next.

Invoices every two weeks

Hours go into your tracker, the invoice comes from GG Solutions Oy. No lock in: we continue for as long as it is useful to you.

Terms

Trial week
25 hours on a bounded task, fixed €2,000 excl. VAT
Rate
€80/h excl. VAT; longer engagements or a fixed scope are agreed separately
Invoicing
GG Solutions Oy, Finnish Business ID 3465139-7, every two weeks by the hour; EU reverse charge where it applies
Contract
IT2018 terms or your own template; confidentiality, and the rights to the code are yours
Capacity
20 or 40 hours a week, usually starting within a week
How
Remote, or hybrid from Helsinki; your repositories, tickets and chat
Languages
English, Finnish and Russian

Questions

Are you a freelancer or a company?

A company: GG Solutions Oy, Finnish Business ID 3465139-7. I invoice hours or an agreed scope. I am not looking for employment.

What do you not do?

Java, .NET, native mobile and PLC programming. The core is Python, PHP and Laravel, React and TypeScript, PostgreSQL and MySQL, Docker and AI integrations.

Can your end client interview you?

Yes. Many end clients want their own interview and a CV with references, and both are fine.

What if it does not work out?

The first week is a bounded task. If the result does not convince you, we stop there and you pay only for the hours worked.

Which time zone?

Central and Northern European working hours are covered. I work remotely, or hybrid from Helsinki.

Tell me about the AI feature that has to reach production

Send the stack, the models and what does not work today. You get my CV and the two closest examples the same working day.

jyri.gromyko2002@gmail.com +358 44 978 2028
Tell me about the AI taskDownload CV (PDF)

Other pages

Full CV and all projectsSubcontracting for software housesPython development as a subcontractorLaravel and React as a subcontractorIntegrations: ERP, banks and POSWooCommerce development for agenciesAI for everyday business work