Architect · Builder · Leader

Designing the operating models and trust infrastructure that make enterprise AI deliver.

VP, AI & Cloud Architecture at Deutsche Bank. I work on two connected problems: the operating models and governance regulated enterprises need to move AI from pilots to measurable impact, and the protocol infrastructure that makes multi-agent AI verifiable. One is an organizational problem, the other an engineering one. In regulated environments you have to solve both.

Read the Enterprise AI Playbook View Research Writing LinkedIn GitHub
Regulated financial services Author · Agentic AI for Serious Engineers Designer · AIP + LDP Published research · arXiv
Sunil Prakash

Flagship Work

The management system for enterprise AI.

Flagship · Interactive Enterprise AI Transformation Playbook Most enterprise AI programs never reach measurable P&L impact. What the ones that do get right: five operating principles, four maturity stages, and the governance, architecture, and measurement that turn capability into business value. 47 pages 5 principles 4 maturity stages 10 min readiness test Read the playbook The framework Take the assessment
Architecture Position Pack Enterprise Data Architecture Data platform architecture for regulated enterprises. Mesh patterns, quality frameworks, lakehouse design, and the operating model that makes data AI-ready. 2026→ Editorial Publication Agentic AI for Serious Engineers A field guide to reliable, evaluable, production-grade agent systems. Lab reports from real builds, field notes on the decisions that matter, and recipes for the patterns that recur in production. Book · Field Notes · Recipes→

Research

The technical foundation beneath enterprise AI governance.

When AI scales to multi-agent architectures, governance needs protocol-level answers. Published research on identity, delegation, provenance, and collective reasoning, with experiment code and open-source implementations.

AIP · arXiv:2603.24775 Agent Identity Protocol In a scan of ~2,000 public MCP servers, none enforced authentication. AIP adds verifiable delegation tokens, with adapters for CrewAI, Google ADK, and LangChain. 0.049ms verification 100% adversarial reject IETF draft-prakash-aip-00→ LDP · arXiv:2603.08852 An Identity-Aware Protocol for Multi-Agent LLM Systems Identity-aware delegation and routing. Six research questions with honest null results on where it helps and where it does not. ~12x latency cut 96% attack detection Interactive paper→ DCI · arXiv:2603.11781 From Debate to Deliberation: Structured Collective Reasoning Typed reasoning moves, preserved disagreement, guaranteed convergence. Honest results on when the cost is and is not justified. +0.95 vs debate 9.56 hidden-profile Interactive paper→

Selected Writing

Enterprise AI strategy, governance, and systems architecture.

AI Leadership Why Most Enterprise AI Programs Fail Before They Start The gap between AI strategy documents and execution systems. Why the organizational share of the problem outweighs the technical one. 2026→ Governance Your ML Risk Framework Wasn't Built for GenAI. Here's What's Missing. Why traditional model risk management fails for LLMs. Hallucination policy, GenAI risk dimensions, deployment gates, and what a complete framework looks like. 2026→ AI Leadership The Middle Management AI Gap AI programs fail at the middle layer. The people who should be supervising AI-augmented workflows were never equipped to do it. 2026→
All writing →

Open Source

Selected infrastructure projects.

Benchmark enterprise-rag-bench RAG patterns benchmarked for enterprise. Five chunking strategies, five retrieval patterns, evaluation harness, guardrails, and observability. github.com/sunilp→ Framework ai-governance-framework AI governance for regulated industries. Risk classification, model lifecycle controls, EU AI Act mapping, and worked examples for credit scoring and fraud detection. github.com/sunilp→
All projects on GitHub →

Background

Experience

  • Deutsche Bank VP, AI & Cloud Architecture — 2021–Present
  • Halialabs Chief Scientist, AI & Platform Architecture — 2018–2021
  • Cognizant Technology Specialist / Software Architect — 2012–2017
  • Syntel Software Engineer / Project Lead — 2007–2012

Education

  • PGPMAX (Executive MBA) Indian School of Business
  • M.Tech, Enterprise Business Analytics National University of Singapore
  • B.E., Computer Science RGTU, India

Contact

Open to leadership conversations.

If you're shaping the next stage of enterprise AI capability, data platform modernization, or technology transformation, I'd welcome the conversation.

LinkedIn Newsletter GitHub sunil@sunilprakash.com