AI Engineering Leader · Founder, Ambharii Labs · Inventor, G-ARVIS · Creator, Bulwark

Anil
Prasad

I build production AI for
regulated industries.

28 years delivering AI/ML platforms in healthcare, energy, and finance. $4.1B in measurable outcomes. Four open-source frameworks in production — most recently Bulwark, the agent-security framework defending AI systems against prompt injection. I write Field Notes: Production AI — bi-weekly dispatches from the inside of regulated AI.

Bi-weekly · #HumanWritten · No spam · Unsubscribe anytime
22 Years in production AI
$3.3B Outcomes delivered
Anil Prasad — Head of Engineering and Product
100 Most Influential AI Leaders
Anil Prasad Head of Engineering & Product
About

Where Deep
Experience Meets
What's Next

I started my AI career at ISRO, training neural networks before transformers existed. What has stayed constant across three decades is a belief that AI systems are only as valuable as their reliability in production — not their performance on a benchmark.

Across Tech, Techstartups, Healthcare, Lifesciences, Energy & Utilities, Insurance, Banking, and Fintech I have led platform transformations that generated over $4B in measurable business outcomes. The work I am most proud of is not the technology — it is the trust that business stakeholders placed in AI systems I built, because those systems told the truth about uncertainty when it mattered.

I co-founded the CAIO Circle Tri-State Chapter to build the executive AI leadership community that I wished had existed when I was navigating these decisions alone. I publish and speak to share what the journey actually looks like from inside enterprise AI — not the demo, the production.

Stanford University
IEEE Member
CAIO Circle Co-Founder
100 Most Influential AI Leaders
IISc Research Background
Forbes Technology Council (Aspirant)
▸
2024 – Present
Head of Engineering & Product
Duke Energy Corporation · CASPAR Platform
▸
2024 – 2024
Head of Software, Products, Data & AI Engineering
Ambry Genetics · Genomics Platform
▸
2021 – 2024
VP Engineering · AI Platforms
R1 RCM · CloudmedAI Integration ($4.1B acquisition)
▸
2011 – 2018
Director - Software Engineering
UnitedHealth Group · Enterprise Data, AI and Analytics Platform
▸
2000 – 2024
Software, Data and AI Engineering Roles · Software, Data, AI/ML Platforms
Xcel Energy . Medtronic · Accenture . Wipro . ISRO/IISC
Platforms

Production AI,
at Scale

Production-grade AI platforms built from real requirements at Ambharii Labs. No demos. No prototypes.

🛡️
Bulwark
Agent Security Framework · Open Source · Just Launched

Production-grade Python framework defending AI agents against prompt-injection attacks. Five-layer defense — Input Sanitizer, ML + pattern Detector, Compartmentalized RBAC, Encrypted Audit Trail, and Human Confirmation Gates. MCP-native and compliance-ready for HIPAA, NERC CIP, and SOC 2.

Agent Security Prompt Injection MCP RBAC HIPAA / SOC 2 Apache 2.0
5layers
Defense Depth
MCP-native
Integration
Apache2.0
License
View Product →
💰
ARIA RCM Platform
Revenue Cycle AI · Live

11 Autonomous AI Agents powering denial prevention and revenue recovery for healthcare revenue cycle management. Built with G-ARVIS Observability and ARGUS Self-Correction to predict claim denials before submission and optimize A/R workflows at scale.

11 AI Agents G-ARVIS Observability ARGUS Self-Correction RCM
11agents
Autonomous AI
Live
Status
Self-correct
ARGUS Engine
View Platform →
🎯
G-ARVIS / Argus-AI
LLM Observability & Scoring · Live

Enterprise LLM observability and scoring platform. Monitors six dimensions of production LLM health — Groundedness, Accuracy, Reliability, Variance, Inference Cost, and Safety — to ensure AI systems perform when stakes are real.

LLM Observability Scoring Guardrails Production Monitoring
6dims
Health Metrics
Live
Status
Entgrade
Scale
View Product →
More platforms — AETHER, PulseFlow, SAM3, GenomixIQ — at ambharii.com →
Framework

The G-ARVIS
Framework

Six dimensions of production LLM health — distilled from building AI systems that govern billions in capital decisions across regulated industries.

Read the Full Article →
G
Groundedness
Output traceability to source. Hallucination prevention.
A
Accuracy
MAPE + ECE. Calibrated confidence, not just correctness.
R
Reliability
P99 latency and SLO adherence under real load.
V
Variance
Semantic stability across identical inputs.
I
Inference Cost
Cost per correct answer, not per API call.
S
Safety
Guardrail calibration and compliance audit trail.
Expertise

Full-Stack
AI Leadership

From data infrastructure to model deployment to business translation — the complete stack of skills required to ship AI that actually works in production.

🤖
AI / ML / Deep Learning
LLM Fine-tuningRAG SystemsAgentic AINLPComputer VisionTime-SeriesForecasting
⚙️
MLOps & LLMOps
Model LifecycleDrift DetectionEvaluation PipelinesFeature StoresCI/CD MLObservability
🏛️
AI Governance
SOX ComplianceHIPAAExplainabilityBias AuditRisk FrameworksSafety
🛡️
AI Agent Security
Prompt Injection DefenseMCP SecurityAgent RBACEncrypted Audit TrailsNERC CIP / SOC 2Red Teaming
Writing

Thought Leadership
From the Trenches

Subscribe All Articles →
Medium May 2026 15 min New
The Web Is Now Weaponized Against Your AI Agents

Prompt injection is no longer theoretical. Live attacks now target MCP servers, agentic browsers, and tool-using LLMs through hidden HTML, Unicode abuse, and role-marker injection. Introduces Bulwark — an open-source five-layer defense framework for production AI agents.

Agent Security Prompt Injection MCP Bulwark Open Source
Medium Feb 2026 18 min
The LLM Metrics That Actually Matter in Production

Why benchmark scores are a distraction — and the 8 measurements that will make or break your AI system when real money is on the line. Introduces the G-ARVIS framework for production LLM observability.

LLMOpsG-ARVIS
LinkedIn Feb 2026 5 min
Why Agentic AI Changes the Evaluation Problem Entirely

Single-turn accuracy metrics break down when your LLM is taking multi-step actions. What needs to change in your observability stack before you ship agentic workflows to production.

Agentic AILLMOps
Medium May 6, 2026 8 min New
Claude Code Authentication: Cross-Platform Setup Script

If you've onboarded a teammate to Claude Code in the last six months, you've probably had this conversation: "It says I'm being charged per token, but I have a Pro subscription?" Six auth methods, one priority chain — and the open-source script I built to solve it across macOS, Linux, Windows, and WSL.

Claude CodeOpen SourceDeveloper Tools
Full archive — 10+ essays on production AI →
New series · Quarterly · Just announced

The Regulated AI
Incident Review.

A quarterly, anonymized post-mortem on production AI failures in healthcare, energy, and finance. Modeled on the NTSB aviation incident report. Aviation has the NTSB. AI has nothing. This series breaks that silence.

Read the program → Submit a case
Issue 01 — Q3 2026
CASE 01
Healthcare RCM agent silently approved $2.4M in non-covered procedures over 6 weeks.
CASE 02
Grid-operations agent issued a maintenance recommendation into a SCADA queue during a heat-wave alert.
CASE 03
FinServ KYC agent answered a prompt-injected adverse-media query with a sanctions clearance it had no source for.
Recognition

Earned in
Production

🏆
100 Most Influential AI Leaders
USA
⚡
IEEE Member
Institute of Electrical & Electronics Engineers
🎓
Stanford University
Alumnus
🌐
CAIO Circle Co-Founder
Tri-State Chapter
Contact

Let's Build
Something
That Matters

Open to conversations about engineering leadership, AI platform architecture, speaking engagements, and advisory roles at companies building AI that has to work when stakes are real.

LinkedIn GitHub Substack
Open to enterprise partnership, strategic acquisition, and Distinguished Engineering leadership conversations.
Contact: anil@ambharii.com · DM on LinkedIn
Anil.Prasad
HEAD OF ENGINEERING & PRODUCT · AI PLATFORM LEADER
LINKEDIN X MEDIUM DEV.TO SUBSTACK GITHUB BULWARK HUGGING FACE HACKER NEWS NEWSLETTER #humanwritten