Madhu Sambangi

🚀 Most production issues aren’t caused by bad code, they’re caused by bad app design. If you’re building applications for the cloud, containers, or microservices, the 12-Factor App is still one of the most practical playbooks out there. Why it matters 👇 ✅ Predictable deployments ✅ Easy scaling ✅ Cleaner config management ✅ Fewer “works on my machine” problems From stateless processes to environment-based configuration, these principles help teams move faster with confidence in production. 📖 I wrote a clear, practical breakdown here on Medium: 👉 https://lnkd.in/ePaNAeBj 💬 Which of the 12 factors do you find hardest to follow in real projects? #12FactorApp #CloudNative #DevOps #SoftwareEngineering #Microservices #BackendDevelopment #ScalableSystems

11

1 Comment

How did we used Domain Driven Design  concept to break down our monolithic system into bounded contexts, and modernize it into cloud-based microservices? Also, how did we implement the bubble context architecture pattern? Read our Engineering Technical blog Credits : Danish Garg https://lnkd.in/dhPSDf4V

36

AWS Hadoop Data Engineer (HDFS to AWS Migration) – Hybrid, New Jersey Lead migration of on‑prem HDFS/Hadoop workloads to AWS cloud. Build and optimize ETL/ELT pipelines using Glue, EMR, Lambda, and S3. Convert legacy Spark/Hive jobs into AWS‑native PySpark workflows. Develop scalable data processing solutions using Python, PySpark, and Spark SQL. Implement data quality, validation, and governance best practices. Work with structured and unstructured data across hybrid environments. Troubleshoot issues across Hadoop clusters, AWS services, and ETL pipelines. Collaborate with architects and cross‑functional teams during cloud modernization.

13

14 Comments

🏥 Automated Prior Authorization with AI Agents — From 3 Days to 4 Hours We recently deployed an autonomous multi-agent system that handles prior authorizations end-to-end — reading clinical notes, validating payer rules, and submitting requests without human intervention.   🚨 The Problem Prior authorization was exhausting our clinicians:   • 72-hour average turnaround   • Manual policy lookup across 50+ payers   • 31% denial rate → constant resubmissions This wasn’t just inefficiency — it was lost clinical time and delayed care.   🤖 The Solution: Multi-Agent Architecture 📄 Document Ingestion Azure Document Intelligence converts clinical notes → structured FHIR resources (observations, conditions, medications) 🧠 Policy Intelligence (RAG Pipeline)   • Coverage policies embedded in Weaviate   • Hybrid search:   • BM25 → exact policy matching   • Vector similarity → semantic understanding   • Nightly payer updates via Playwright automation   🧩 Agent Orchestration (LangGraph) Four specialized agents collaborate:   • Clinical Context Agent — interprets patient FHIR data   • Policy Matching Agent — retrieves coverage criteria   • Gap Analysis Agent — identifies missing documentation   • Submission Agent — submits to payer portals/APIs   🔧 Agent Tools (via MCP)   • Custom FHIR MCP server (Epic / Cerner integration)   • ICD-10 / CPT validation lookup   • Payer portal automation (Playwright)   • Attestation document generation   🧬 Reasoning Layer Claude Sonnet performs clinical reasoning:   • Understands medical necessity   • Interprets payer language   • Maps findings → coverage criteria ⚙️ Workflow Orchestration Prefect manages scheduling, retries, and SLAs   🔒 Compliance & Explainability Every agent decision is written to an immutable audit log with reasoning — supporting HIPAA compliance and clinical transparency.   📊 Results After 6 Months ✅ 94% auto-approval rate (up from 69%) ✅ 31% reduction in denials ✅ 4-hour turnaround (down from 72 hours) ✅ Clinicians regained 15 hours/week   🧰 Tech Stack LangGraph • Claude Sonnet 4 • Weaviate • Azure Document Intelligence • Prefect • Playwright • Custom MCP servers (FHIR & coding)   💡 Key Insight Prior authorization isn’t a document processing problem. It’s a multi-step reasoning problem across clinical context, policy rules, and workflows. Single models struggle. Multi-agent systems with domain-specific tools succeed. #GenAI #SRE #IncidentResponse #PagerDuty #LangGraph #LLMOps #PlatformEngineering #AIEngineering #Observability #SARSoftwareInc    Learn more about our Success:   https://lnkd.in/e7N3Xgew Learn more about our expertise: https://lnkd.in/db_Mzi96   

41

3 Comments

FABCON 2026 in Atlanta was nothing short of mesmerizing-where bold ideas, breakthrough tech, and real-world impact came together seamlessly. An electrifying blend of vision, innovation, and execution that set the tone for the future of data and AI. Now, it’s a wait-and-watch moment to see how much of it proves truly real, scalable, and competitively differentiated. ** AI Foundry: Fabric IQ is positioned as a capability that makes data AI-consumable for agents and AI apps via OneLake ** OneLake Transform transformations are applied on top of shortcut or mirrored data (zero‑copy, zero‑ETL), preserving the original source while exposing a transformed view in OneLake. New Connectors have been added (SharePoint in preview). ** Data and Catalog mirroring / zero copy options across open table formats from snowflake and Databricks into Fabric ** SAP Fabric and Salesforce Fabric version with zero copy and write back functionality ** Availability of CoPilot across various platforms within Fabric exclusively designed to support that platform only ** Fabric AI and Data Agents ** Zero ETL Data Mirroring onto Onelake from Azure data for PostgreSQL. Analytics on Operational workloads made simple. ** Building declarative AI Pipelines made simpler with dspy and pixeltable Thanks to Cognizant leadership for the opportunity. Rajesh SreepadaSatish Hegde FABCON & SQLCON - The Microsoft Fabric & SQL Community Conferences Microsoft Fabric

23