RAG Development Services.
Retrieval augmented generation systems that let your team ask questions of your own documents, contracts and data, and get answers with citations.
Retrieval augmented generation, or RAG, is how you make an AI answer from your documents instead of from the internet. EqualPixels builds RAG systems and internal knowledge assistants so your team can ask a question in plain language and get an answer that cites the exact policy, contract or specification it came from.
What is RAG?
RAG works in two steps. First the system searches your own content for the passages most relevant to the question. Then it passes those passages to a language model and asks it to answer using only that material. The model does not recall your data from training, it reads it at the moment of asking. The approach was introduced in the 2020 research paper Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks and is now the standard way to build an AI knowledge base.
That is why RAG is the default architecture for internal assistants: it keeps answers current, keeps them attributable, and keeps your documents out of anyone’s training set.
RAG vs fine-tuning: which do you need?
| RAG | Fine-tuning | |
|---|---|---|
| Best for | Facts, policies, documents that change | Style, format, consistent task behaviour |
| Updating | Re-index, minutes | Retrain, days |
| Citations | Yes | No |
| Cost to start | Low | High |
If the question is “what does our documentation say”, the answer is RAG. Fine-tuning is for shaping how the model responds, not for teaching it facts. Most projects that ask for fine-tuning need retrieval instead.
What we build
- Internal knowledge assistants
- and enterprise RAG over policies, SOPs, wikis, past proposals and technical documentation, so staff can chat with your documents with permissions applied and see only what they should
- Contract and document question answering
- , including clause lookup and comparison across versions
- Customer-facing support search
- grounded in your help centre, with citations shown to the user
- Engineering and product assistants
- over codebases, specifications and ticket history
- Data-aware assistants
- that combine document retrieval with live queries against your database
How we build a RAG system that is actually accurate
- Audit the corpusDuplicated, contradictory and outdated documents are the number one cause of wrong answers. We find them before indexing.
- Chunk with structure, not by character countSplitting on headings and sections preserves meaning that naive chunking destroys.
- Hybrid retrievalKeyword (BM25) and vector search over embeddings together, then a reranker, because pure vector search misses exact identifiers, product codes and names.
- Permission-aware indexingAccess control applied at retrieval, so an answer can never leak a document the user cannot open.
- Evaluate honestlyA fixed question set with approved answers, scored for retrieval hit rate, answer correctness and citation accuracy, run on every change and traced in Langfuse so a regression is visible before users see it.
- Show the sourceEvery answer links to the passage it came from, so people can verify rather than trust.
How much does a RAG system cost?
RAG development typically starts at $8,000 for a knowledge assistant over one defined document set, with permissions, citations and an evaluation set included.
- Number of sources
- SharePoint, Google Drive, Confluence, a help centre and a database each need their own connector and sync.
- Content quality
- Scanned PDFs, tables and contradictory versions take preparation before indexing.
- Permissions
- Mirroring per-user access from the source systems adds the most engineering.
- Accuracy target
- A higher bar means a larger evaluation set and more tuning of chunking and reranking.
Running cost is mostly model usage per question plus vector storage. pgvector inside your existing PostgreSQL adds almost nothing; a managed vector database is a monthly fee.
The RAG stack we build on
- Vector databases
- pgvector when you already run PostgreSQL, Pinecone or Qdrant when the corpus or query volume calls for a dedicated vector database
- Embeddings and reranking
- OpenAI and open-weight embedding models, chosen by how they score on your own questions, with a reranker over the top results
- Models
- OpenAI, Anthropic Claude, Gemini or open-weight models depending on data residency, deployed in your cloud where regulation requires it
- Orchestration
- LangGraph or plain TypeScript and Python pipelines, with structured outputs so citations come back as data rather than prose
- Access
- a private chat interface, an API, or an MCP server so ChatGPT, Claude and your agents can query the same index
- Evals and LLMOps
- evaluation sets in CI, tracing in Langfuse, and alerts when retrieval quality drifts after content changes
Case study
Prozilio: AI proposals for freelancers
Prozilio, our AI proposal tool, stores each persona’s skills as embeddings in PostgreSQL with pgvector and writes proposals that cite the freelancer’s own projects rather than a generic template.
Read the case study: Prozilio: AI proposals for freelancers
Frequently asked questions
Still have questions? Write to us at [email protected]
How do I connect ChatGPT to my company’s internal data?
Not by uploading everything into a chat window. You build a retrieval layer over your documents and expose it either through a private assistant or through a Model Context Protocol server that ChatGPT and Claude can call. That keeps access controlled and answers current.
What is RAG and when should I use it instead of fine-tuning?
Use RAG when the answer lives in documents that change. Use fine-tuning when you need a consistent output style or a narrow repeated task. They are complementary, and RAG is almost always the cheaper first step.
Is our data safe?
Your documents stay in your storage. We use enterprise endpoints that exclude data from training, and where regulation demands it we run open-weight models entirely inside your own infrastructure.
How accurate is it?
Accuracy is a property of your content as much as the model. We measure it against a fixed evaluation set and report the number rather than describing it as high. If accuracy is unacceptable, the fix is usually the source documents, not the model.
How long does a RAG build take?
A working internal assistant over a defined document set typically takes three to six weeks, with most of that time spent on content preparation, permissions and evaluation rather than on the retrieval code itself.
How much does RAG development cost?
RAG projects typically start at $8,000 for a knowledge assistant over one defined document set, with permissions, citations and an evaluation set included. The price rises with the number of source systems, the state of the content, per-user permissions and the accuracy you need. Model usage and vector storage are running costs on top.
Can it work across several systems?
Yes. We routinely index SharePoint, Google Drive, Notion, Confluence, a help centre and a database together, with a single permission model across all of them.
Let’s build what’s next.
Tell us about your RAG Development Services project. We’ll reply within 24 hours with a clear plan and estimate.