arun.v

About

Senior Full Stack and AI engineer. 12+ years shipping production SaaS.

I specialize in multi-agent orchestration, agentic safety, and LLMOps for regulated, high-scale environments. I build the unglamorous parts of platform: the typed state object, the citation verifier, the eval pipeline, the integration that has to survive a duplicate webhook at 3am. I prefer measured outcomes over adjectives, contracts over conventions, and a small surface area over a clever one.

Current work: an enterprise intelligent document processing platform at Staples, a LangGraph classify, extract, validate, escalate pipeline over Azure AI Document Intelligence OCR and GPT-4o extraction, with confidence-scored routing and human-in-the-loop exception queues.

Recent work: PolicyMind, a four-agent RAG platform with 92.5% citation accuracy and 100% escalation recall behind CI-enforced RAGAS gates; CollectMind, a LangGraph vehicle telemetry policy engine with COVESA VSS validation and confidence-gated deployment; and Agentix ERP, a greenfield multi-tenant SaaS ERP with a 10-step autonomous onboarding agent on AWS Bedrock.

Before that: 11+ years across healthcare SaaS. EMR clinical workflows, e-prescribing on AWS with DEA EPCS-compliant identity federation, and a multi-source EHR migration program that moved millions of patient records with zero data loss.

I hold an MS in Computer Science and Engineering from Oakland University and a BS in Computer Science from Jawaharlal Nehru Technological University. I am based in the United States and open to relocation and remote work. Open to senior backend, full-stack, and AI engineering roles. H-1B with approved I-140, transfer-eligible.

Where I have spent my time, what I owned, and what shipped.

  1. May 2026 to present

    Senior AI Engineer

    Staples (Remote)

    • Intelligent Document Processing platform: LangGraph classify, extract, validate, escalate pipeline over a Python / FastAPI microservices backend. Azure AI Document Intelligence OCR feeds GPT-4o structured extraction against Pydantic-typed schemas, with per-field confidence scoring and human-in-the-loop exception queues.
    • Multi-source ingestion (email, SFTP, scan/OCR feeds, supplier portals) into an event-driven Kafka flow, validated against vendor master data and two-way and three-way PO matching before reaching finance systems.
    • Evaluation harnesses on Databricks with MLflow tracking: field-level precision and recall, straight-through-processing rate, and model drift enforced as CI release gates. Docker on Kubernetes with Azure DevOps CI/CD.
  2. Apr 2021 to Mar 2026

    Senior Software Engineer

    Modernizing Medicine (ModMed)

    • PolicyMind (Sep 2025 to Mar 2026): internal compliance RAG platform. Four-agent LangGraph system with hybrid retrieval (Qdrant + BM25 + RRF + cross-encoder), NLI-based hallucination verification, 92.5% citation accuracy at p95 < 800ms, 100% escalation recall, 167 CI-gated tests.
    • EMA clinical workflows and mmRxService (Mar 2025 to Sep 2025): Python / FastAPI e-prescribing backend on AWS (Lambda, SQS/SNS, Redis, PostgreSQL) with an Amazon Bedrock-backed clinical decision support panel and a LangGraph prior-authorization drafting workflow behind human review gates. Okta and SureScripts integration.
    • Denial triage and appeal automation (Jan 2024 to Oct 2024): GPT-4o + LangGraph pipeline over TriZetto 835 ERA files with a compliance-guard citation verifier. 100% canary recall (20 of 20 fabricated citations blocked), 96% of citations verified verbatim, denial review time cut from 20-40 minutes to 3-5 minutes per case.
    • EHR data migration (Apr 2021 to Mar 2026): TRAKnet, Sammy, and Exscribe to EMA. Hundreds of practices, millions of records, zero data loss across all production cutovers, with Snowflake-based cross-practice compliance reporting.
  3. Dec 2014 to Apr 2021

    Software Developer / Senior Software Developer

    Nemo Health LLC

    • NemoRx (May 2017 to Apr 2021): multi-tenant SaaS e-prescribing with React frontend, Cosmos DB partition-key isolation, SureScripts integration, ELK-backed audit dashboards, and DEA EPCS-compliant identity federation via Okta and Exostar.
    • TRAKnet (Dec 2014 to Apr 2021): podiatry EHR platform on .NET / WinForms across hundreds of active practices, with TriZetto clearinghouse, ANSI 837/835 EDI, and PCI DSS payment integration. Led distributed offshore engineering team.
  4. May 2014 to Sep 2014

    Software Developer

    ATCO Industries Inc.

    • Enterprise .NET applications and SQL Server data models supporting operational workflows.
  5. 2013 to 2015

    MS, Computer Science and Engineering

    Oakland University, Michigan

  6. 2008 to 2012

    BS, Computer Science and Engineering

    Jawaharlal Nehru Technological University, India

Skills

Stack

Day-to-day work and working knowledge, grouped honestly.

Agentic AI / GenAI

Day-to-day

  • LangGraph
  • LangChain
  • Multi-Agent Orchestration
  • ReAct Tool-Calling
  • Versioned Prompt Registry
  • GPT-4o (structured output)
  • text-embedding-3-large
  • NLP (NLI, Text Classification, Semantic Search)
  • Qdrant
  • BM25 + Vector Hybrid + RRF
  • Cross-Encoder Reranking
  • RAGAS
  • LLMOps
  • Intelligent Document Processing (Azure AI Document Intelligence, OCR + LLM Extraction)

Working knowledge

  • Model-Agnostic Routing
  • A/B Testing

AI-Assisted Development

Day-to-day

  • Claude Code
  • Parallel Multi-Agent Coding Workflows
  • Spec-Driven Agent Scaffolds (ADRs, state snapshots, task backlogs)
  • CI-Gated Agent Output
  • Prompt-Driven Code Generation and Review

Agentic Safety & LLM Security

Day-to-day

  • NLI Hallucination Verification
  • PII Redaction (referential)
  • Confidence-Gated Deploy
  • Action Authorization Registries
  • Compliance-Guard Architecture

Working knowledge

  • LlamaGuard

Backend

Day-to-day

  • C# / .NET Core
  • ASP.NET Core Web API
  • Java / Spring Boot
  • Python / FastAPI
  • Entity Framework
  • REST
  • Swagger / OpenAPI
  • Microservices
  • Event-Driven Architecture

Working knowledge

  • gRPC
  • WCF (legacy)

Frontend

Day-to-day

  • TypeScript
  • React
  • Angular 17 (RxJS)
  • Next.js 15
  • Tailwind CSS
  • SPA Architecture

Working knowledge

  • WPF / UWP / XAML (legacy)

Clinical / Pharma Domain

Day-to-day

  • HL7, FHIR, SMART on FHIR
  • Clinical Decision Support (CDS)
  • CCD / C-CDA
  • ANSI 837/835 EDI
  • ICD-10 / CPT
  • NPI / DEA / EPCS
  • SureScripts
  • HIPAA / PHI / PII Compliance

Working knowledge

  • BAA-Governed Integrations

Security

Day-to-day

  • OAuth2 / JWT
  • IAM Least-Privilege Roles
  • DEA EPCS Two-Factor Identity (Okta / Exostar)
  • HIPAA PHI Isolation
  • PCI DSS Payment Integrations
  • Secure API Design

Cloud & Deployment

Day-to-day

  • Azure AI Foundry
  • Azure (App Insights, Cosmos DB)
  • AWS (Bedrock, Lambda, SQS/SNS, DynamoDB, S3, SES)
  • Docker
  • Kubernetes
  • GitHub Actions
  • Azure DevOps

Working knowledge

  • Vercel

Data / Streaming

Day-to-day

  • PostgreSQL
  • TimescaleDB
  • SQL Server / T-SQL
  • Cosmos DB
  • MySQL
  • Qdrant
  • Apache Kafka
  • Redis
  • Elasticsearch / Kibana
  • Snowflake

Working knowledge

  • Databricks

ML / Data Science

Day-to-day

  • Scikit-learn
  • Isolation Forest
  • Time-Series Anomaly Detection
  • Feature Engineering
  • Model Training / Tuning / Eval

Observability & Practices

Day-to-day

  • Application Insights
  • Langfuse
  • Prometheus
  • Grafana
  • RAGAS CI gates
  • xUnit
  • Moq
  • Eval-gated deploys

Working knowledge

  • SLSA
  • OWASP ASVS

Outside the editor

Not all of it is code

Cricket

Captain of the Oakland County Cricket Club. Team management and field captaincy turn out to be uncannily similar to technical leadership: the people problem is the problem.

Designing Agentic AI applications

I have been designing agentic AI applications to solve problems where multi-step reasoning, tool use, and autonomous decision-making produce better outcomes than traditional code.