Back to home

Research

Research & technical papers

Original work on the architecture of production-grade AI systems for regulated environments — RAG under GDPR and the EU AI Act, sovereign LLM deployment, and the engineering economics of compliance.

2026·33 pages·Insightrix Working Draft·EN

Compliance-Aware Retrieval-Augmented Generation for Regulated Financial-Reporting Corpora

A Real-World Evaluation on SEC EDGAR Filings

CARAG is a five-stage RAG architecture that treats compliance as a first-class property of the index, the retriever, the generator, and the audit log. Evaluated on a benchmark built from 6,000 real SEC EDGAR filings (26,595 chunks across seven recent quarters), it cuts the constraint-violation rate from 81.12% to 0.00% and the output-disclosure rate from 21.29% to 0.00%, at a Token-F1 cost of only 4.8 points and 0 ms of 95th-percentile latency overhead.

CVR 81.12% → 0.00%ODR 21.29% → 0.00%4.8 F1 cost0 ms p95 latency overhead

Aru Bhardwaj

Fractional CTO architecting sovereign AI systems for startups and scale-ups across Europe. Custom ML, agentic RAG, and secure LLM infrastructure. 7+ years turning complex data into production intelligence.

Malt
Upwork

Contact

Services

  • Fractional CTO & AI Strategy
  • MVP Development & Rapid Prototyping
  • Sovereign LLM Deployment (OVHcloud, Scaleway)
  • Multi-Cloud AI (AWS Bedrock, Vertex AI, Azure)
  • RAG Pipelines & Autonomous Agents
  • GDPR & EU AI Act Compliance
  • Generative AI & Prompt Engineering
  • Machine Learning & Predictive Analytics

Monthly playbook

Practical AI essays for founders and tech leaders. One email a month.

Tactical AI essays, monthly.

© 2026 Insightrix SASU. All rights reserved.Aru Bhardwaj, Fractional CTO & AI Strategist

60 Rue François Ier, 75008 Paris, France · SIRET 989 236 856 00013 · TVA FR42989236856