New York City Metropolitan Area
I build AI that survives contact with the real world. Over 20 years I've led engineering and product through three platform transformations tied to more than $4B in measurable outcomes. My work lives where most AI dies. The gap between a demo that impresses the room and a system that runs every day inside a regulated business, with real money and compliance on the line. That gap is where a Forward Deployed Engineer lives, and it's how I work best. I embed with the people who own the problem, write the code, own the outcome, and don't leave until it's in production and trusted from Fortune 10 to Fortune 200 to startups, scaling teams from 25 to 200+ engineers. I build in the open, and my frameworks run in production: - AgentMesh: A governance proxy in front of every AI tool a team uses: 85% cache hits, 75% lower cost, zero code changes. Shipped on PyPI, Docker, and Hugging Face. - ARIA: eleven autonomous agents fixing healthcare revenue cycle in production. A working agentic system, not a copilot. - Bulwark (agent security vs. prompt injection), PulseFlow (ML lifecycle), JARVIS (AI observability), SAM3 Perception Hub (vision on Meta's SAM), TorchForge (governed PyTorch). My edge sits at three intersections: 1) AI-native architecture: AI as a first-class component: evaluation, observability, safety, and the governance that ships under HIPAA, FDA, and GxP. 2) Modernization at scale: transformation without feature freezes: AWS/Azure/GCP migrations, monolith to microservices, zero-downtime data moves, and API-first strategies behind multibillion-dollar acquisitions. 3) Teams that ship: hire for potential, build psychological safety, measure outcomes over output, grow engineers from mid-level to principal. I'm an inventor on multiple patents and a Top 100 Most Influential AI Leader in the USA (2024). I write Field Notes: Production AI: for engineers who have to make this work for real. The engineering world is splitting into people who use AI as a tool and people who architect AI as a system. I build the second kind platforms where AI and traditional engineering do what neither could alone. If you're taking AI from pilot to production at scale, or building the team that will, let's talk. GitHub: github.com/anilatambharii · Writing: medium.com/@anilAmbharii · More: anilsprasad.com
Leading the engineering product transformation for Duke Energy's CASPAR (Cost Allocation Schedule Performance Analytics and Reporting), modernizing renewable energy cost analytics across Fortune 150 operations processing $50B+ annual cost data. Managing 20+ engineers with hands-on technical leadership driving AWS-native platform modernization. Engineering Leadership & Execution: Leading 20+ engineers: code reviews, PR approvals, design specifications, roadmap definition Driving strategic technical architecture for renewable energy analytics transformation Collaborating with engineers, compliance, security teams influencing enterprise designs Establishing engineering excellence culture with code quality standards and rapid delivery Platform Architecture & AWS Modernization: Architecting AWS migration (S3, Lambda, RDS, Redshift) replacing legacy Excel workflows Implementing Airflow orchestration for complex data pipelines and build automation Building event-driven microservices supporting renewable energy cost allocation schedules AI/ML Innovation & Prototyping: Architecting GenAI capabilities using OpenAI GPT-3.5 and BERT for intelligent automation Designing RAG workflows for automated insight generation Prototyping agentic frameworks with FastAPI/Celery for autonomous task execution Evaluating hybrid vector search (FAISS, Elasticsearch) with NLP preprocessing (spaCy, NLTK) Establishing MLOps evaluation frameworks with BLEU/ROUGE metrics and Datadog observability Strategic Value: Modernizing mission-critical renewable energy analytics for Fortune 150 operations Enabling C-suite data-driven decision-making through scalable analytics Building foundation for AI-native operations unique in utilities sector Establishing technical standards and governance for enterprise-wide adoption Tech Stack: AWS (S3, Lambda, RDS, Redshift), Airflow, Python, FastAPI, Celery, OpenAI GPT-3.5, BERT, FAISS, Elasticsearch, spaCy, NLTK, Datadog, Power BI
Open-source AI/ML engineering and technical thought leadership translating Fortune 150 production experience into frameworks the whole community can use. AI governance, MLOps, and agentic systems hardened against $20B+ in enterprise data processing and $4B+ in business outcomes. Flagship work: - AgentMesh: a governance proxy that sits in front of every AI tool a team uses (Claude Code, Copilot, ChatGPT, Gemini, and custom agents). Three-layer semantic cache, per-team token quotas, cheapest-model routing, and an Ed25519-signed audit log, plus a Chrome extension. 85% cache hits and 75% lower cost with zero code changes. Apache 2.0 shipped on PyPI, Docker Hub, and Hugging Face. - ARIA: eleven autonomous agents fixing healthcare revenue-cycle management in production (patented). A working agentic system, not a copilot. - GenomixQA: multi-agent genomics platform selected for a World Economic Forum cohort. - Bulwark (agent security vs. prompt injection), PulseFlow (production MLOps — deployment, monitoring, drift detection, governance), JARVIS (AI observability), SAM3 Perception Hub (computer vision on Meta's SAM), TorchForge (governed PyTorch). Enterprise-to-open-source translation: - MLOps, RAG, and agentic patterns drawn from leading 60+ engineers across Fortune 150 (Duke Energy, Ambry Genetics, R1 RCM). - AI governance for regulated industries: FDA, HIPAA, GxP, model ethics, production reliability. - Distributed-systems playbooks: zero-downtime migrations, dual-write strategies, event-driven architectures. Recognized as a Top 100 Most Influential AI Leader in the USA (2024). I write Field Notes: Production AI, and publish deep-dives on LinkedIn and Medium. Tech: Python, PyTorch, FastAPI, LangGraph, CrewAI, sentence-transformers, HuggingFace, Kubernetes, Docker, AWS, Azure, GCP, Vector Search, Airflow, Datadog.
Co-founded the Tri-State chapter of CAIO Circle, a peer community connecting Chief Data & AI Officers across NY/NJ/CT region. Facilitate roundtable discussions on enterprise AI challenges including production-scale deployment, AI governance, model operationalization, and talent architecture. Active AIM Council member contributing to industry dialogue on AI leadership. Established founding membership at Microsoft NYC launch event (January 2026). Skills: AI Adoption at Scale, Artificial Intelligence (AI) [+18 skills]
Led engineering transformation at Ambry Genetics, architecting zero-downtime MySQL to Vitess migration with GenAI-powered clinical intelligence. Managed 60+ engineers, 20+ data scientists across product, clinical, compliance, delivering genomics infrastructure processing millions of variant records. Engineering Leadership: - Architected zero-downtime cloud migration eliminating performance bottlenecks threatening clinical SLAs and regulatory compliance - Led 60+ engineers with C-suite alignment balancing innovation, compliance, clinical quality - Established AI governance framework addressing FDA, HIPAA, GxP requirements Technical Architecture & GenAI: - Designed sharded Vitess enabling 10x query performance (<100ms), optimized for ML and LLM integration - Architected GenAI platform using SageMaker for automated annotation and LangChain agents for clinical decision support - Implemented event-driven architecture with AWS Lambda, S3, real-time streaming for ML feature extraction - Built dual-write shadow traffic migration with consistency validation, automated rollback, canary deployment - Orchestrated blue/green deployments with agentic monitoring for auto-reconciliation AI/ML at Scale: - Operationalized ML pipelines for automated annotation, phenotype analysis in clinical workflows - Deployed LangChain agents for clinical querying, evidence synthesis, compliance automation - Established MLOps with continuous retraining, A/B testing, drift detection Business Impact: - 99.99% uptime, 10x performance, 50% clinical review cycle reduction via GenAI automation - Migrated millions of records with zero data loss - Enabled M&A readiness contributing to Tempus AI acquisition - Halved incident workload through self-healing systems Tech Stack: Vitess, MySQL, AWS (SageMaker, Lambda, S3, ECS/EKS), GenAI (LangChain), Python, Go, Datadog
Led global engineering transformation for R1 RCM's CloudmedAI platform, multi-tenant healthcare analytics solution. Directed 100+ engineers across software, data, security through post-M&A consolidation and platform modernization. Engineering Leadership & M&A: - Led 100+ global team through multi-cloud consolidation post-acquisition - Orchestrated migration from fragmented on-prem, AWS, Azure into unified Azure E5 tenant - Managed diverse stacks (PHP, Java, .NET, Ruby, RPA) establishing unified architecture - Drove C-suite alignment balancing innovation, compliance, operational stability Platform Architecture: - Architected CloudmedAI: data extraction/ingestion, Contracts & Rules Engine, Azure ML orchestration, analytics - Built multi-tenant infrastructure supporting thousands of users processing millions of transactions - Designed distributed data platform using Spark, Databricks, Synapse for real-time analytics - Implemented microservices with event-driven workflows, API-first design Security & Compliance: - Established enterprise Security Program from ground up - Remediated 50,000+ vulnerabilities implementing IAM, DevSecOps, WAF, API hardening - Achieved PCI DSS, HIPAA, HITRUST compliance - Built security-first culture with automated scanning, continuous monitoring AI/ML & GenAI: - Pioneered enterprise GenAI using ChatGPT Turbo 3.5 and Pythia APIs for operational intelligence - Integrated ML pipelines for predictive analytics, claim processing, intelligent routing - Established MLOps with model governance and monitoring Business Impact: - Enabled $4.1B Cloudmed acquisition through platform consolidation and security remediation - Delivered platform serving thousands of users with enterprise-grade reliability - Transformed fragmented post-M&A landscape into unified healthcare analytics platform Tech Stack: Azure, AWS, Spark, Python, Databricks, Synapse, Terraform, MongoDB, Postgres, React, .NET, Java, Airflow, OpenAI GPT-3.5