I operate distributed infrastructure and build systems that explain their own behavior.
At work, that means Multi-cloud CPaaS infrastructure across 10+ regions · 1M+ calls/day · 99.9% availability.
Outside work, I apply infrastructure ideas—observability, caching, framing, congestion control, reconciliation, and failure isolation—to systems that do not normally receive infrastructure-level discipline.
This is not a manually maintained profile page. It is a generated view of declarative engineering state.
OPERATOR Harsh Daga
ROLE Cloud Infrastructure Engineer
DOMAIN Multi-cloud CPaaS infrastructure
PRODUCTION 10+ regions · 1M+ calls/day · 99.9% availability
PRINCIPLE boring systems, explicit failure modes
SYSTEM MODEL
├── provision Terraform · Terragrunt · Ansible
├── reconcile Kubernetes · Helm · Argo CD
├── observe Prometheus · Grafana · ELK · Thanos
├── communicate SIP · RTP · distributed networking
└── experiment agent observability · LLM transport · incident automation
SYSTEM: cairn
class agent observability
objective make agent behavior explainable
mechanism traces → fingerprints → outcomes → improvement
properties local-first · measurable · reproducible
state active
SYSTEM: lattice
class LLM transport
objective reduce provider and token inefficiency
mechanism compression → cache → framing → flow control
properties provider-neutral · protocol-oriented
state active
SYSTEM: incidentscribe
class incident automation
objective remove low-judgment on-call toil
mechanism detection → triage → timeline → postmortem
properties human escalation · explicit handoffs
state active
SYSTEM: catalyst
class event intelligence
objective surface signal from noisy public events
mechanism ingest → extract → evaluate → alert
properties evidence-oriented · infrastructure-minded
state active
2026-07-14Cairn — Release Cairn 1.1.1 after PyPI filename rejection (#35)2026-05-26Lattice — Merge pull request #22 from Harsh-Daga/refactor/phase-14-transport-consolidation2026-01-01Catalyst — chore: remove unnecessary documentation files2025-12-14IncidentScribe — Merge pull request #6 from Harsh-Daga/doc-update
Change feed is reconciled from public repositories; the last committed state remains visible when the API is unavailable.
- boring technology
- visible failure
- reproducible systems
- local-first design
- narrow responsibility
- Operating regional telephony infrastructure across AWS, Azure, and GCP.
- Building causal observability for coding agents through Cairn.
- Exploring protocol-level efficiency for LLM traffic through Lattice.
- Writing down failure modes usually discovered only during incidents.