AI strategy & discovery
A short, structured look at where AI pays off in your business — and where it does not. You get a prioritised roadmap, effort estimates and an explicit no-go list.
Rīga, Latvia · working across the EU
Your AI partner in Latvia — from idea to production.
Smitcom Projects designs, builds and runs AI systems that hold up in daily use. Strategy first, working software second, and someone still on the hook after launch.
- LLM applications
- Retrieval & RAG
- Process automation
- Data platforms
- MLOps
Services
What we do
Five engineers' worth of opinions, one senior team. We take projects from the first honest conversation about whether AI is the right tool, through to a system your people actually use.
LLM applications & assistants
Customer-facing and internal assistants grounded in your own data, with evaluation, guardrails and cost control built in from the first sprint — not bolted on later.
Knowledge & retrieval
Documents, tickets, contracts and wikis turned into a searchable, citable knowledge layer, so answers come with a source your team can check.
Process automation
Document intake, back-office workflows and integrations with the systems you already run. Humans stay in the loop exactly where the risk sits.
Data platforms
Pipelines, storage and clean data models — the unglamorous groundwork that decides whether the AI layer above it is any good.
Run & improve
Monitoring, evaluations, model upgrades and spend control after go-live. AI systems drift; someone needs to own that.
Selected work
Things we build
A sample of the problems companies bring us. Details of specific engagements available on request.
Logistics & trade
Document intake that reads itself
Incoming invoices, waybills and customs paperwork parsed, classified and routed into the ERP. Anything the model is unsure about goes to a human queue instead of quietly guessing.
Professional services
An assistant that knows the archive
Years of proposals, reports and contracts made searchable in plain language, with citations back to the source document so nothing has to be taken on faith.
Operations
The boring workflow, automated
A repetitive daily process — read, decide, record, notify — moved into a monitored pipeline, with a dashboard showing exactly what it did and what it skipped.
Approach
How we work
Small steps with a decision point at the end of each one. You can stop after any of them and still own something useful.
-
01
Discover
We map the process, the data and the constraints, then rank the candidates by value against effort. Ends with a written recommendation — sometimes that recommendation is "don't build this".
-
02
Prototype
The riskiest part gets built first, on your real data, measured against a test set we agree on up front. No demo-ware.
-
03
Ship
Production build: integrations, access control, logging, fallbacks and a rollout plan that lets people ease into it rather than wake up to it.
-
04
Operate
Evaluations keep running, costs stay visible, models get upgraded deliberately. Hand-over to your team whenever you are ready for it.
About
A small team that stays close to the work
Smitcom Projects is an AI consultancy and development studio based in Rīga. We work directly with the people who own the process — no account layer between you and the engineers building your system.
We work in Latvian and English, keep data inside the EU by default, and are happy to start with a fixed-scope pilot before anyone commits to anything larger.
- Based inRīga, Latvia
- Working withLatvia, Baltics & EU
- LanguagesLatvian · English
- DataEU residency by default
- EngagementsFixed-scope pilots · retainers
Because we care
Because we care about our planet's future
AI runs on electricity. Every model we pick, every pipeline we schedule and every job left running overnight has a footprint — so we treat efficiency as part of the engineering rather than someone else's problem.
-
Right-sized models
The smallest model that clears the bar, not the largest one available. It is cheaper for you and it burns less power.
-
European infrastructure
Workloads run in EU regions — close to your data, under EU rules, and on comparatively clean grids.
-
Less waste
Batch what can wait, cache what repeats, and switch off what nobody is using. Most of the savings are simply not running things twice.
Latvia is more than half forest. We would like to keep it that way.
Contact
Tell us what you're trying to fix
A rough description is enough to start. We'll reply with an honest read on whether it's worth building and what it would take.
- Email io@smitcom.lv
- Where Rīga, Latvia