Website builds and maintenance
New pages, content updates, schema markup, accessibility fixes, and performance tuning for client sites.
Let’s talk Internal platform / AIOS
AIOS is the internal platform that runs Marketing & AI’s technical work. It coordinates site builds, SEO and AEO rounds, content QA, and client reporting across a local server mesh: multiple machines on our office network, not a rented cloud. Every task passes through human review before it reaches a client.
What is AIOS? An AI operating system built by Marketing & AI to run its own agency operations. It handles the technical side of client work: builds, audits, reporting: so the team can focus on strategy and client relationships.
01 / Workload
Each of these used to be manual. Now they run through the system with a person checking the output.
New pages, content updates, schema markup, accessibility fixes, and performance tuning for client sites.
Crawl audits, structured data, internal linking, answer-readiness checks, and search console analysis.
Every piece of generated output gets reviewed against the client brief, brand rules, and factual accuracy before it ships.
Automated data pulls, performance summaries, and recommendations: delivered on schedule without manual assembly.
02 / Architecture
AIOS runs on hardware we own and operate. Client data stays on our network.
We run a multi-node server mesh across our office. The system routes tasks to the right machine based on what each job needs: GPU capacity for model inference, fast storage for large crawls, dedicated resources for long-running agents. When a job needs stronger reasoning than our local models can provide, it routes to the Claude API.
03 / Research
AIOS powers a quarterly research crawl of 2,474 Houston business websites.
The Houston AI Visibility Index measures how well local businesses appear in AI-powered search: ChatGPT, Perplexity, Google AI Overviews, and Bing Copilot. AIOS runs the full pipeline: crawling every site, analyzing structured data, testing answer readiness, and compiling the results into a public report.
04 / Claude integration
Claude Code already handles all of our site work. The API is next.
We use Claude Code for every website build, update, and maintenance task across our client portfolio. The next phase moves the rest of the AIOS workload: code generation agents, document analysis, and quality assurance: from local open-source models onto the Claude API for the jobs where reasoning quality directly affects client outcomes.
A support question may use different words from the policy that answers it. Document chunking keeps each rule near its exceptions. Hybrid search combines exact identifiers with semantic search, then reranking can reorder the candidate passages.
For a test question with known sources, recall at k measures how many of those sources reached the shortlist. Groundedness asks whether the answer’s claims are supported by what was retrieved. These checks catch different failures.
An embedding stores a numerical representation of a passage. A vector database can index those values using HNSW. When combining matches from BM25 and vector similarity, reciprocal rank fusion combines their positions. Monitor embedding drift after the document population changes.