Expertise · LLM integration

Language models as a dependable part of your software

Classifying documents, extracting data from emails, summarising reports. A language model can take over a lot of manual work, provided you treat it as a component that sometimes errs, and design your software accordingly.

What a good integration looks like

  • Structured output. The model returns data in a fixed schema that your software can validate, rather than free text.
  • The right model for each task. A small, fast model for simple work and a larger one where it matters. That often accounts for most of the savings.
  • Resilient to abuse. User input can try to steer the model. Good integrations account for that.
  • No lock-in. An abstraction layer lets you switch providers, or move to a model running in the EU or on your own infrastructure.
What that looks like in code
llm/extract.pyPython 3.12
from datetime import date
from decimal import Decimal
from pydantic import BaseModel, Field

# fixed schema: your software validates what the model returns
class Invoice(BaseModel):
    supplier: str
    invoice_date: date
    total: Decimal = Field(ge=0)

async def extract(text: str) -> Invoice:
    # small model where possible, large model where needed
    model = "small" if len(text) < 4_000 else "large"
    return await llm.parse(text, output=Invoice, model=model)

Request profiles

Profiles of suitable LLM developers within 48 hours

  • No obligation: you only pay for hours worked
  • Dutch-speaking, from our own network and assessed by us
  • On-site, hybrid or remote

Indicative € 106–132 per hour, all-in, excl. VAT

Rather call? +31 35 7111 373

What we select for

  1. 01

    Cost-aware

    Understands how tokens, caching and model choice drive the bill, and manages accordingly.

  2. 02

    Robust

    Catches errors and nonsense output before it reaches a customer or another system.

  3. 03

    Pragmatic

    Chooses plain code where that works better than a language model.

Technology we often see

Tools change fast. We select on principles, but this is what the developers in our network use in practice.

PythonPydanticFastAPIOpenAIAnthropicMistralOllamaLiteLLM

Profiles we are looking for

Need a Python developer for this?

Send us your vacancy or request. Within 48 hours you hear which developers in our network fit and what it costs.