AI Accountability · Policy Analysis · System Investigation

Auditing the systems
that shape decisions.

I help organizations understand, audit, and navigate complex systems at the intersection of artificial intelligence, law, and public policy.


What I do

End-to-end research, analysis, and advisory across the intersections of AI, law, data, and public accountability.

01

AI System Auditing

Evaluate LLM reliability, bias, and failure modes. Design duty-of-care frameworks and conduct comparative model analysis across deployment contexts.

02

Policy & Legal Analysis

Analyze legislation, court cases, and regulatory systems. Translate legal complexity into actionable insights for technical and policy audiences alike.

03

Data & Text Analysis

Build reproducible pipelines for large-scale document analysis — entity extraction, sentiment, co-occurrence networks, and targeting pattern detection.

04

Information Integrity

Audit misinformation dynamics, trace narrative propagation, and design evidence-based communication strategies grounded in empirical methodology.


AI-Augmented Structural Analysis
for Policy, Research, and High-Stakes Documents

I audit complex documents to reveal what they are actually doing beneath the surface.

Not summaries. Not opinions. Structural, data-backed analysis of language, intent, and internal consistency.

Using AI as an analytical instrument—not a content generator—I surface the signals that traditional review misses.

Hidden Contradictions

Internal inconsistencies that undermine stated positions or expose unstated assumptions.

🔀
Competing Ideologies

Conflicting worldviews coexisting within a single document — often unacknowledged.

📢
Rhetorical Pressure

Measurable signals of urgency, threat, and authority used to steer interpretation.

🗺
Narrative Shifts

Where tone, framing, or intent changes across sections — and what that signals.

🔍
Coordination Signals

Patterns of language that suggest coordination, fragmentation, or deliberate manipulation.

Why This Matters

Most high-impact documents today are too large to critically read in full, present themselves as internally consistent, and rely on rhetoric that goes unmeasured and unchallenged. Traditional analysis misses this.

I make the invisible structure visible — and defensible.

  • Are too large to critically read in full
  • Present themselves as internally consistent
  • Rely on rhetoric that goes unmeasured and unchallenged
A human-led, AI-augmented audit pipeline
01
Framework Design

Define analytical dimensions tailored to the document — power, legitimacy, urgency, sentiment, and more.

02
Computational Analysis

Process entire documents at scale — thousands of text segments — without sacrificing precision.

03
Multi-Model Validation

Cross-check outputs across systems to reduce bias and error. No single-model dependency.

04
Structural Mapping

Identify topic clusters, narrative shifts, and internal fractures across the document body.

05
Synthesis & Reporting

Deliver clear, evidence-backed findings — not raw data. Structured for both technical and non-technical stakeholders.

Core Deliverables

  • Executive Report — clear, defensible findings
  • Section-Level Analysis — where shifts and contradictions occur
  • Topic & Narrative Mapping
  • Rhetorical Scoring — power, urgency, legitimacy
  • Identified "Fault Lines" within the document

Optional Outputs

  • Raw datasets (.csv / .xlsx)
  • Visual dashboards
  • Cross-document comparison — competing policies or organizations
🏛
Policy & Legislative Analysis

Identify internal contradictions and ideological splits in regulatory and legislative texts.

📣
Think Tank / Advocacy Audits

Detect narrative framing and strategic messaging in advocacy and policy documents.

🏢
Corporate & Regulatory Review

Surface inconsistencies in disclosures, filings, or strategic positioning documents.

🔬
Research Integrity & Bias Detection

Analyze tone, framing, and structural bias in academic and institutional work.

What Makes This Different

Most AI work today generates content. This does the opposite. Every result is human-directed, cross-validated, and evidence-based — with no automation without oversight, no black-box conclusions, and no single-model dependency.

Oversight
No automation without human direction
Transparency
No black-box conclusions
Reliability
No single-model dependency

You are not buying AI output.

You are hiring a Principal Investigator using AI as a high-resolution analytical instrument.

Clarity is leverage. Structure is truth.

In an environment flooded with generated content and unverified claims, I provide both.


How I work

01

Methodology-first

Every conclusion is traceable and defensible. Pipelines are built for reproducibility from the start, not retrofitted.

02

Institutionally independent

No employer, no funder, no conflicts of interest. Analysis goes where the evidence leads.

03

Designed for scrutiny

Code, data, and reasoning are structured so others can replicate, audit, and challenge the work.

04

Clear communication

Complex systems become actionable insight. Reports are written for both technical and non-technical stakeholders.


Tools & capabilities

Research infrastructure built for rigor, reproducibility, and scale.

Python pandas spaCy NetworkX Plotly pdfplumber regex text processing LLM evaluation prompt engineering Jupyter notebooks GitHub Windows CLI Bash / Linux entity extraction sentiment analysis co-occurrence networks targeting-pattern detection

Work with me.

Available for document audits, research projects, and advisory engagements.

contact@narrovue.com