AI your auditor will sign off on.
Bedrock RAG (the model answers from your own documents) that cleared a Tier-1 bank’s OSFI B-13 review in six weeks. If your data can’t leave your VPC, we’re built for you.
At a glance
- Build
- 6–16 weeks
- then a 3-week handoff
- Data locality
- Stays in your VPC
- Built for
- OSFI B-13 · SOC 2 · HIPAA
- First deliverable
- 48 hours
A production RAG system, inside your VPC.
In your AWS account, with PII masking and a full audit trail.
- RAG system in production, in your AWS account
- PII masking before any model call
- Eval harness you run on every change
- Audit trail of every prompt and response
- Cost controls: per-team budgets, token limits and circuit breakers
Four stages, from use case to evaluation.
- 01Use-case scoping and model selectionAn honest conversation about what AI can and can’t do for your team: Bedrock or self-hosted, Claude or Llama. We tell you when the answer is a well-placed SQL query.
- 02RAG architectureChunking, embeddings and hybrid search tuned to your documents. Financial filings, engineering manuals and legal contracts each need different retrieval. We don’t ship defaults.
- 03PII detection and data isolationPre-processing that detects and masks 23+ PII entity types before a token reaches the model, using Comprehend, custom regex or both. Reversible only for authorized workflows.
- 04Guardrails, audit and evaluationBedrock guardrails for content policy, CloudTrail and CloudWatch for every prompt and response, and a test suite your team runs against every model change.
Built for VPC-bound data by default.
Compliance is reviewed before code is written.
An enterprise AI platform cleared OSFI B-13 in six weeks, with no data leaving the bank.
Canadian Tier-1 bank, 2024 to present
Not a fit if you…
- Just need a chatbot on your website. We build the backend and will refer you for the UX.
- Can happily use the OpenAI API and don’t need data locality. We’re overkill and overpriced for that.
- Want us to train a foundation model from scratch. That’s not what we do.
Questions about models and risk.
How much does an AI engagement cost?
Scoped RAG systems run $80–180k and full agent platforms with complex integrations $200–400k. Compliance-heavy builds (OSFI, HIPAA) add 20–30% for the audit evidence work.
Why Bedrock over OpenAI or Anthropic direct APIs?
For regulated clients: data locality, VPC endpoints and no data leaving your environment by default. For everyone else we use whichever model serves the use case, and we are not paid by AWS to recommend it.
Can you help us decide if we need AI at all?
Yes, and we have told clients they did not, on the first call. Sometimes a well-placed SQL query is the answer.
What about hallucinations?
We evaluate against your actual retrieval corpus, not a generic benchmark, and report a confidence score per answer. When retrieval quality drops below a threshold the system fails closed: "The system doesn't know" is a valid answer.
Do you work with open-source models?
Yes, Llama, Mistral and fine-tuned variants via SageMaker when the use case calls for it. Most production RAG systems need better retrieval, not fine-tuning.
Tell us what you're building.
We reply within one business day.
First call: 30 minutes, on a Tuesday or Thursday, with two engineers, not a sales rep. If we’re the wrong fit, we’ll say so and name a firm that is.