Skip to main content
Synonet. Synonet · Hamburg

Process · How we work

Our
process.

Structured, transparent and tailored to your needs — from the first idea to ongoing operations.

Analysis · Prototype
Implementation · Training
Operations · Optimisation
Our approach — 04 From analysis to optimisation
01

Analysis & Discovery

We start with an intensive workshop to understand your business processes, data sources and goals. Which data should be used? Which compliance requirements apply? Where are the biggest opportunities?

02

Proof of Concept

Based on the analysis we build a first prototype — usually with a relevant open-source model and real (anonymised) company data. The PoC shows within three to four weeks what is possible.

03

Implementation & Training

After the go-ahead we implement the solution in your infrastructure: model optimisation, building the RAG pipeline, fine-tuning and integration into your existing systems.

04

Operations & Optimisation

After go-live we don't disappear — quite the opposite. The best AI deployments emerge from continuous monitoring, feedback loops and regular model updates. We work closely with your team.

Frequently asked questions — FAQ Questions we hear almost every time
How long does a typical LLM project take?
That depends heavily on scope and complexity. A proof of concept can be ready in 3–4 weeks. A complete production system with a RAG pipeline and fine-tuning typically takes 8–16 weeks.
What hardware is required?
That depends on the model and usage intensity. For a 7B model, 1–2 NVIDIA RTX 4090s or an A100 with 40 GB are often enough. We are happy to advise you on hardware selection and, on request, procure it for you as well.
Can we keep using an existing cloud LLM in parallel?
Yes — many of our customers run a hybrid strategy: sensitive queries locally, less critical tasks remain in the cloud. We help you find the right split.
What happens when the model needs to be updated?
We take over update management within the scope of our support contract. Model updates are validated in a test environment before they go into production — without downtime for your team.

Non-binding first consultation — free of charge

Ready for the first step?

Book a consultation