Hello, I'm

Inderdeep Singh

+--------+
  > _    
         
         
+--------+

Building reliable systems as a

Passionate about infrastructure automation, cloud-native technologies, and operational excellence.

Technical Skills

Technologies and tools I work with

Infrastructure

AWSLinuxOn-PremiseNetworking

Containers & Orchestration

KubernetesDocker

Languages

GoPythonJavaScriptBash

Databases

PostgreSQLRedisMongoDB

DevOps & CI/CD

TerraformGitHub ActionsJenkinsAnsible

Observability

PrometheusGrafanaOpenTelemetry

Featured Projects

Some of the systems and tools I've built

DNS-SD Service Discovery

_tcp.srvvmagentblackboxvm-storevmalerticingaDNS SRV recordsscrape targetsprobe endpointstime-series storerule evaluationon-call alertsdiscoversprobeswritesevaluatesnotifies
Fig. 1Targets are discovered from DNS SRV records rather than a service registry — the hop in red is the point of the system, and the only one drawn in colour.

Implemented DNS SRV-based service discovery to monitor infrastructure outside Consul. Enables dynamic discovery of streaming nodes through vmagent, blackbox exporter probes, and VictoriaMetrics—closing monitoring gaps for edge infrastructure.

PrometheusVictoriaMetricsDNSMonitoring

StableXchange

client-restgrpc-serverpostgres-ledgerkafka-eventsmongo-walletsclickhouse-logswallet-servicetransfer requestvalidation, routingdouble-entry rowsdurable event logbalancesanalytics sinkidempotent consumerrequestscommitsemits eventsships logsreads / writesconsumessynchronousasynchronousKEY
Fig. 2The ledger commit is the only synchronous write in the system. Everything downstream is asynchronous, so a stalled analytics sink or a lagging consumer cannot fail a transfer.

Event-driven payment system for INR ↔ stablecoin transfers. Built with Go, gRPC, Kafka, and a multi-database architecture (MongoDB, Postgres, ClickHouse) featuring reconciliation patterns, ledger safety, and idempotent webhook handling.

GogRPCKafkaPostgreSQLMongoDB

Cylinder Misfire

vehicle-ecubrowserdiagnose-apiclaudepgvectorOBD-II · ISO 15765-4Web Bluetooth · ELM327FastAPI · asyncpgranked root causescomplaints · TSBs · recallslive datascanpromptgrounds
Fig. 3The reasoning step is grounded before it runs: retrieved complaints, bulletins and recalls for that specific vehicle go into the prompt, so the model is citing the corpus rather than recalling it.

AI diagnostic platform for independent auto shops, built with a working mechanic on the thesis that diagnosis is the part of the job nobody pays for. A technician live-scans a vehicle over Web Bluetooth or types the codes in by hand, and gets back ranked root causes with the verification steps to prove or kill each one — grounded in real complaints, service bulletins and recalls for that exact vehicle rather than whatever the model happens to remember.

Next.jsFastAPIPostgreSQLpgvectorClaude APIWeb Bluetooth
// contact info
socials:
├── github
├── linkedin
└── x/twitter