Proof, Architecture & Work Record

See the work, the reasoning, and verifiable structure

This page demonstrates how Taghi Molavi and InTen engineer search discoverability and answer-engine presence across Google, ChatGPT, and AI systems. Every section links to an architectural build, open benchmark, or verifiable evidence pattern.

What this page is for

Understand three things: the core problem, the system built, and what can be checked in public. Client identities and confidential business metrics remain strictly private.

From work to evidence
ProblemBuildEvidence

Problem → Build → Evidence

Case Studies & Architectural Impact

Specific client identities remain confidential under NDA; the problem pattern, system build, and qualitative evidence are detailed below.

CASE STUDY [01]National Digital Publisher & News Media

Context

High-frequency editorial publication with hundreds of daily articles and distributed syndication channels.

Problem

Rapid publishing cadence outpaced structured metadata hygiene, causing fragmented entity signals in Google News and near-zero source attribution in generative answer engines.

Build & Architecture

Architected automated JSON-LD NewsArticle and Author entity graphs, established canonical author disambiguation via Wikidata/sameAs references, and built pre-publication schema validation pipelines.

Evidence

Qualitative outcome: Achieved consistent Knowledge Graph entity attribution and validated Rich Results across all primary editorial sections. Verified schema compliance with zero syntax errors on Search Console indexation queues.

🔒 Client identity and internal proprietary CMS configurations remain confidential under non-disclosure agreements.

CASE STUDY [02]B2B Industrial Equipment & Engineering Brand

Context

Specialized machinery and industrial manufacturing firm serving domestic and export markets.

Problem

Extensive product catalogs lacked hierarchical topical clustering. Generative search engines hallucinated technical specifications or attributed product capabilities to competitors.

Build & Architecture

Engineered multi-layered Product and Service schema models, rebuilt semantic specification tables for machine extraction, and aligned reciprocal multilingual hreflang clusters across regional markets.

Evidence

Qualitative outcome: Established unambiguous entity boundaries in Google Knowledge Panels and eliminated model hallucinations across standard commercial evaluation prompts.

🔒 Commercial contract values and specific corporate entities are strictly protected under confidentiality agreements.

CASE STUDY [03]Multilingual Professional Services & Consulting Firm

Context

Cross-border consultancy handling international client inquiries across three regional languages.

Problem

High volume of unqualified, unstructured leads created operational bottlenecks while primary service pages lacked semantic depth for AI retrieval.

Build & Architecture

Constructed an integrated Digital Twin intake workflow with interactive screening criteria, connected validated brand knowledge bases, and structured localized service landing pages.

Evidence

Qualitative outcome: Disqualified mismatched inquiries before human consultant intake, established validated service schema across all locales, and achieved reliable retrieval in targeted answer engine queries.

🔒 Customer inquiry logs and proprietary diagnostic protocols remain confidential.

Open-Source Tools & Repositories

Three open-source tools on GitHub that make the measurement methodology and evaluation logic transparent.

TypeScript · Benchmark

geo-scope

Problem: There was no repeatable way to compare how an entity appeared in answer engines.

Built: A benchmark that records prompts, answers, citations, and retrieval differences.

Example: Example: run the same question across engines and inspect where the brand and its sources appear.

Verifiable result: Result: a reproducible report and method, rather than a vague visibility claim.

View on GitHub ↗
MCP · Python

mcp-agent-skills-hub

Problem: Agents had skills scattered across projects with unclear quality.

Built: A registry and evaluation layer for finding the right reusable skill.

Example: Example: before acting, an agent can discover and inspect a search or validation skill.

Verifiable result: Result: skills become discoverable, reusable, and reviewable.

View on GitHub ↗
n8n · Docker

n8n-agent-skills

Problem: n8n automations could work in a demo but lacked production validation.

Built: Reusable skills with explicit checks before and after an agent action.

Example: Example: validate links, structure, and the API response before publishing content.

Verifiable result: Result: an inspectable workflow instead of blind execution.

View on GitHub ↗

Why open source?

Open source means part of the tool and method can be inspected on GitHub. It does not mean client information is published; it means the work is not supported by marketing claims alone.

If you do not know how your brand appears in AI answers, start here.

Request a project review →
Verifiable Evidence & Architectural Impact | Taqi Molavi