/work
Backend Engineer, IBM Cloud · May 2024 to present
ETL Health Monitoring Microservice
IBM Client and Partner Success Award, June 2025 · OpenTelemetry, Grafana
BEFOREEngineers checked collector health by opening a UI and reading one percentage. There were no alerts and no detail by region, zone or cluster.
AFTERA service tracks 50+ ETL collectors, sends timely alerts and shows status per region, with metrics visualized in Grafana.
Collector health dropped below the 92% threshold about twice a week on average, so each drop used to depend on someone happening to look. Alerting reduced undetected pipeline failures.
decisions made
One health number or per-region status
CHOSEPer-region and per-cluster status, with alerts.
WHYA single percentage says that something is wrong, not where. Engineers needed the where.
OpenTelemetryGrafanaalertingobservability
Decision Engine pre-production
Event-driven automation with RAG on a Granite LLM · I built the engine component
The engine takes multi-source inputs such as ServiceNow incidents, Slack messages and anomaly detection, standardizes them, chooses a strategy based on severity and category, and runs actions: health checks, notifications and root cause suggestions.
In pre-production, one flow ran end to end: a NodeNotReady alert from the anomaly detection tool triggered the self-healing system, which remediated the node.
decisions made
One collector or several
CHOSESeveral.
WHYA single collector is a single point of failure. I argued for this in the design discussion, then implemented it.
Kafka or RabbitMQ
CHOSENeither, for now.
WHYPre-production load didn't justify running another piece of infrastructure. It stays in the backlog.
event-drivenRAGGraniteServiceNow
Self-Healing System
Ansible · ServiceNow · Slack · 10+ playbooks
Ansible playbooks for the incidents that come up most often. They run when an alert fires instead of waiting for someone to pick the incident up and they can also be started from Slack and through APIs. The system is integrated with ServiceNow.
The automation cut toil and mean time to repair by 50%. In the Decision Engine's pre-production run, the Decision Engine used this system as a remediation tool to fix the node.
AnsibleautomationServiceNowSlack
More at IBM
- Developed Python, FastAPI and TypeScript microservices that connect ServiceNow, Slack and internal REST APIs to improve incident triage.
- Led a cross-team review of alert coverage across 20+ VPC services against the golden signals (latency, errors, saturation, traffic) to standardize alerting.
- Maintained a shared Grafana platform used by compute, network, storage and control plane teams, including upgrades and onboarding.
- Containerized and deployed 100+ services with Docker, Kubernetes (IKS) and Jenkins.
/skills
Languages
PythonJavaScriptTypeScriptNode.js
AI and LLM
RAGGranite LLMLangChainOllamaMLOpsGenAI assistants
Cloud and containers
IBM Cloud (VPC)AzureAWSDockerKubernetes (IKS)
Automation and CI/CD
TerraformAnsibleJenkinsGitGitHubJIRAArtifactory
Observability and data
OpenTelemetryGrafanaPrometheusServiceNowRabbitMQMongoDB
CertificationsHashiCorp Certified Terraform Associate (003) · ChatGPT Prompt Engineering for Developers (DeepLearning.AI) · Google Cloud Generative AI Leader (in progress)
