AI advisory
Unsure where AI will help, what it will cost, or what happens to your data? Work directly with me to choose a useful first project.
- Find the work worth improving
- Clarify privacy, cost & ownership
- Agree on a pilot with useful outcomes
Hi, I'm Dinh.
Practical AI.
Personally led.
Your team needs answers, not more files to search. Your sensitive data needs clear boundaries. I help businesses turn these problems into practical AI projects, from direct advice to delivery with a team.
Direct advice from me. Development delivered with a team.
Local Knowledge AIAsk your business documents
Who approves a new vendor?
Answer from your documents
The procurement owner starts the review. Legal reviews the terms, then the operations lead gives final approval.
Vendor handbook.pdfp. 12
Vendor approval flow
- 01Procurement ownerStart the review
- 02Legal reviewReview the terms
- 03Operations leadGive final approval
What we can work on
Start with a real problem. Give the technology a clear job.
Local Knowledge AI
Important answers are buried in a growing archive, but your documents are too sensitive for an unrestricted AI tool. Let's plan a private knowledge system around your business.
- Find answers across scattered documents
- Keep data within agreed boundaries
- Plan for a growing knowledge archive
AI development
Repetitive work and disconnected tools slow your team down. I lead a delivery team to build AI into the workflow you already use.
- Reduce repetitive handoffs & manual work
- Connect knowledge to everyday tools
- Keep people in control of key decisions
Local Knowledge AIYour knowledge is growing.
Your knowledge is growing.
Your data needs to stay private.
The answer may already be in your documents. Finding it means searching folders, opening PDFs, or asking the one person who knows. Let's make that knowledge easier to use, with privacy and access rules defined from the start.
Your private environment
Office AI Box or your own cloud account
Your business documents
Files, scans, tables & diagrams
Local Knowledge AI
Read, organize & find the answer
Knowledge stays hereAnswers for your team
With sources they can check
Sensitive data needs boundaries.
Contracts, customer records, and internal decisions need a clear home. Define where data can be processed and who can see it before choosing an AI tool.
More files. Harder to find answers.
Policies, reports, scans, and diagrams pile up across folders. Help your team ask a business question and find the relevant source, without opening file after file.
Your experts become the search engine.
The same questions keep reaching the same people. Make shared knowledge easier to use, so your specialists can focus on decisions that need their judgment.
Start with your data boundaries: which documents must stay in the office? Which can run in your own cloud account? Who should have access? We use those answers to scope where processing happens, including offline operation when required.
Facing terabytes of files or millions of documents?
A large archive needs a plan beyond a chat interface. We assess what your team needs to find, which records they can access, and how the collection changes. Start with a representative slice, then scope storage, processing, and rollout around the results and your budget.
How we approach your document system
Technical capabilities to scope around your requirements
Complex PDFs
Parse text, scanned pages, and mixed layouts while keeping connections to the original document.
OCR & visual understanding
Extract text from scans and use local vision models to interpret images, charts, and diagrams.
Answers with evidence
Combine keyword and semantic search, rerank results, and link answers to source pages. Flag missing evidence.
Large document collections
Batch ingestion, background processing, progress tracking, retries, and index updates as documents change.
Controlled access
Knowledge-base permissions enforced during retrieval, with access and administrative audit logs.
A self-hosted AI pipeline
Plan parsing, OCR, vision, embeddings, retrieval, answers, storage, and logs within your office or cloud environment.
A question. An answer.
The source behind it.
A number buried under grouped headers. A component identified across two drawings. An answer spread across a faded scan. Explore the source material your team needs to make sense of.
Public document excerpts and fictional samples, with prewritten answers and manually marked reading guides. This preview does not run an AI model.
Local Knowledge AIIllustrative examples
Research archive · Multi-level table
An analyst needs the right number from a dense report, without mixing up tasks, units, or footnotes.
Your question
At a 36-inch work surface, what are the right-slap capture times, and what does the number in parentheses mean?
Answer with a source
In Table 2, follow the 36-inch row into the YTsr group (right slap). The mean is 7.2 seconds, the median is 7 seconds, and the standard deviation is 2.1 seconds. The parenthesized 1 is the interquartile range, as explained below the table.
- Mean
- 7.2 s
- Median
- 7 s
- Standard deviation
- 2.1 s
- Interquartile range
- 1 s
What to check
Connect the grouped YTsr header, its three subcolumns, the 36-inch row, the task legend, and the footnote. A single cell without that context is easy to misread.
Open original PDF (opens in a new tab)
Theofanos et al. (2006), Effects of Scanner Height on Fingerprint Capture, NISTIR 7382. Republished courtesy of the National Institute of Standards and Technology.
A clear path from idea to pilot
Know what we're testing, and what success looks like.
Understand the problem
Start with your team, workflow, data, and the outcome that would make a difference.
Define a useful pilot
Agree on representative documents, access rules, infrastructure, and questions to test.
Test against real work
Can your team find a useful answer, check its source, and respect access boundaries? Test those questions before expanding.
Plan deployment & support
Agree on a rollout, operating costs, ownership, and support that your business can sustain.
One point of contact.
A team when you need to build.
Dinh Van · Independent AI consultant
DinhNV is my personal home on the web. I work independently as an AI consultant, so you work directly with me on strategy, discovery, and technical direction.
When a project moves into AI development, I bring together a team for the agreed scope and stay involved in planning and review. Team roles, deliverables, and support are defined for each engagement.
For Local Knowledge AI, I help you scope a private document system and the infrastructure it needs. An office AI machine or your own cloud account can be part of the plan, even if you're starting without existing infrastructure.
Let's start with your problem.
What slows your team down?
Hard-to-find answers, sensitive documents, or a growing archive? Tell me where work gets stuck, how much data you have, and what must stay private. We can work out the technology and deployment together.