Adith Manikonda Project / 2026 / Systems / Backend
archAIc Reliability Engineering Platform
Distributed reliability engineering platform with observability, metrics, tracing, and anomaly detection.
Source ↗The Problem
Observability tooling is almost always demonstrated against a healthy system, which tells you nothing about whether it catches the failures you actually get at 3am.
The Approach
Four instrumented services with a real dependency graph, where product and payment both depend on auth and db, so a single degraded dependency produces a genuine cascade rather than an isolated error. Failures are injected probabilistically, time-bound and intensity-scaled, so the same experiment can be repeated. An Isolation Forest detector proxies Prometheus metrics and webhooks an operator service that attempts recovery. A shared trace_id follows a request across every hop, which is what makes root-cause attribution possible at all.
Measured
4
Instrumented services
auth, db, product, payment, plus detector and operator
4
Load profiles
k6: normal, spike, endurance, stress
What It Does Not Do
Recovery is executed by an LLM operator rather than a deterministic runbook, so its actions are suggestions with side effects, not guarantees. The chaos layer and the detector are the parts that hold up under repetition.
Built With
- Kubernetes
- OpenTelemetry
- Grafana
- Python
- Observability
More from Adith Manikonda
- 2025 / Systems / BackendSCOTLAND YARDReal-time multiplayer backend supporting 100+ users with persistent game state and server-authoritative gameplay.
- 2026 / Systems / BackendHYBRID SYMBOL TABLECache-conscious, arena-backed symbol table in C11 with adaptive per-scope storage and constant-time scope teardown.
- 2026 / Systems / BackendAETHERQUERYCross-database analytical query engine supporting DuckDB/PostgreSQL/MySQL with adaptive execution.