Hi, I'mTarunTarun

Solutions architect and DevOps lead for an edge-compute AI surveillance platform sustaining 99.9% uptime across 21 live telemetry feeds over 90+ days, after GPU benchmarking and provisioning NVIDIA RTX 4000 servers and PtP/PtMP wireless networks. Runs YOLOv8/TensorRT inference engines with PostgreSQL and MediaMTX (RTSP/RTMP) containerized in Docker Compose, with Zero-Trust enforced through a split-tunnelled WireGuard VPN, strict egress, and Pi-hole DNS sinkholes.

Codified a zero-public-IP Azure landing zone in Bicep (4 zoned subnets, deny-all NSGs, Private Endpoints), eliminating 100% of standing credentials and reducing the attack surface to 0 reachable public assets, validated by Microsoft Defender. Operates an OIDC-federated GitOps pipeline promoting SHA-tagged containers across 30+ deployments a month. Migrated 2.8 TB of on-prem SharePoint data to M365 with 0 data loss using a throttling-aware, multi-threaded Python engine (MSAL, REST APIs) that recovers gracefully from 429/503 tenant responses, alongside MuleSoft API and Intune MDM integration.

Resolved 26 security-audit findings across 37 endpoints and servers by architecting a default-deny WireGuard VPN, deploying Wazuh SIEM, and automating UFW and Bash rules with Ansible for SSH GeoIP anomaly blocking. Cut release cycles 73% (45 to 12 minutes) across 6+ production apps by standardizing Jenkins and GitHub Actions pipelines with SonarQube quality gates, tuning Linux kernels via sysctl, and engineering Restic encrypted backups.

Led an ELT-over-ETL data platform scaling to 24.7M time-series records, designing a resilient 3-tier PostgreSQL warehouse with star-schema modelling across 10+ sources, building delta-load transformers, and managing a Next.js/Plotly monorepo. Reached 99.9% execution success on data-acquisition pipelines by orchestrating headless-browser automation (Selenium/Chromium) through multi-profile Docker Compose and running 19+ idempotent SQL migrations.

Reduced commit-to-production lead time 80% for an AI algorithmic-trading platform by orchestrating a React/C#/Python microservices stack on Azure Kubernetes Service with Helm templates and Azure Pipelines. Engineered horizontal scalability with HPA autoscaling and namespace-scoped RBAC, configured Kubernetes CronJobs for Apache Arrow Flight workloads, and deployed cache-busted Docker builds via Azure Container Registry.

Cut logging-infrastructure costs 80% and removed cluster overhead for 5 GB/day of logs by retiring Elasticsearch and building a lightweight full-text search and trace-correlation engine on SQLite FTS5 and FastAPI. Achieved a validated 24-hour RPO and 2-hour RTO for Supabase/PostgreSQL through a nightly automated dump-and-restore pipeline that strictly verifies schema, role, and sequence integrity.

Reduced p99 API latency 86% (850 to 120 ms) for a healthcare ecosystem by migrating .NET and SQL microservices into a single-tenant Azure App Service Environment (ASE v3) behind Front Door, WAF, and APIM, guided by KQL telemetry. Closed 100% of VAPT findings on the first re-test by mitigating LUCKY13 CBC timing attacks with TLS 1.3, enforcing HSTS, centralizing secrets in Key Vault, and provisioning Mirth Connect with Azure Data Factory SSIS.

Drove MTTR down 65% by engineering a proactive, self-healing observability pipeline that uses CloudWatch telemetry to trigger SNS alerts and event-driven Lambda functions for automated EC2 remediation. Achieved a sub-5-minute RPO in live recovery drills by architecting a highly available multi-tier AWS environment with parameterized CloudFormation, least-privilege IAM, and isolated RDS with point-in-time recovery.

Trimmed cloud observability spend 45% and drove MTTD down to 2 minutes by replacing Azure Application Insights with a self-hosted ELK, Prometheus, and Grafana stack, including custom C# Serilog middleware for the Elastic APM cutover. Accelerated BI releases from 2 hours to under 10 minutes with modular Jenkins declarative pipelines, Apache Superset provisioning, and a custom Python CLI that promotes dashboards with zero standing credentials.

Solutions architect and DevOps lead for an edge-compute AI surveillance platform sustaining 99.9% uptime across 21 live telemetry feeds over 90+ days, after GPU benchmarking and provisioning NVIDIA RTX 4000 servers and PtP/PtMP wireless networks. Runs YOLOv8/TensorRT inference engines with PostgreSQL and MediaMTX (RTSP/RTMP) containerized in Docker Compose, with Zero-Trust enforced through a split-tunnelled WireGuard VPN, strict egress, and Pi-hole DNS sinkholes.

Codified a zero-public-IP Azure landing zone in Bicep (4 zoned subnets, deny-all NSGs, Private Endpoints), eliminating 100% of standing credentials and reducing the attack surface to 0 reachable public assets, validated by Microsoft Defender. Operates an OIDC-federated GitOps pipeline promoting SHA-tagged containers across 30+ deployments a month. Migrated 2.8 TB of on-prem SharePoint data to M365 with 0 data loss using a throttling-aware, multi-threaded Python engine (MSAL, REST APIs) that recovers gracefully from 429/503 tenant responses, alongside MuleSoft API and Intune MDM integration.

Resolved 26 security-audit findings across 37 endpoints and servers by architecting a default-deny WireGuard VPN, deploying Wazuh SIEM, and automating UFW and Bash rules with Ansible for SSH GeoIP anomaly blocking. Cut release cycles 73% (45 to 12 minutes) across 6+ production apps by standardizing Jenkins and GitHub Actions pipelines with SonarQube quality gates, tuning Linux kernels via sysctl, and engineering Restic encrypted backups.

Led an ELT-over-ETL data platform scaling to 24.7M time-series records, designing a resilient 3-tier PostgreSQL warehouse with star-schema modelling across 10+ sources, building delta-load transformers, and managing a Next.js/Plotly monorepo. Reached 99.9% execution success on data-acquisition pipelines by orchestrating headless-browser automation (Selenium/Chromium) through multi-profile Docker Compose and running 19+ idempotent SQL migrations.

Reduced commit-to-production lead time 80% for an AI algorithmic-trading platform by orchestrating a React/C#/Python microservices stack on Azure Kubernetes Service with Helm templates and Azure Pipelines. Engineered horizontal scalability with HPA autoscaling and namespace-scoped RBAC, configured Kubernetes CronJobs for Apache Arrow Flight workloads, and deployed cache-busted Docker builds via Azure Container Registry.

Cut logging-infrastructure costs 80% and removed cluster overhead for 5 GB/day of logs by retiring Elasticsearch and building a lightweight full-text search and trace-correlation engine on SQLite FTS5 and FastAPI. Achieved a validated 24-hour RPO and 2-hour RTO for Supabase/PostgreSQL through a nightly automated dump-and-restore pipeline that strictly verifies schema, role, and sequence integrity.

Reduced p99 API latency 86% (850 to 120 ms) for a healthcare ecosystem by migrating .NET and SQL microservices into a single-tenant Azure App Service Environment (ASE v3) behind Front Door, WAF, and APIM, guided by KQL telemetry. Closed 100% of VAPT findings on the first re-test by mitigating LUCKY13 CBC timing attacks with TLS 1.3, enforcing HSTS, centralizing secrets in Key Vault, and provisioning Mirth Connect with Azure Data Factory SSIS.

Drove MTTR down 65% by engineering a proactive, self-healing observability pipeline that uses CloudWatch telemetry to trigger SNS alerts and event-driven Lambda functions for automated EC2 remediation. Achieved a sub-5-minute RPO in live recovery drills by architecting a highly available multi-tier AWS environment with parameterized CloudFormation, least-privilege IAM, and isolated RDS with point-in-time recovery.

Trimmed cloud observability spend 45% and drove MTTD down to 2 minutes by replacing Azure Application Insights with a self-hosted ELK, Prometheus, and Grafana stack, including custom C# Serilog middleware for the Elastic APM cutover. Accelerated BI releases from 2 hours to under 10 minutes with modular Jenkins declarative pipelines, Apache Superset provisioning, and a custom Python CLI that promotes dashboards with zero standing credentials.