p95.sql
−64%
SQL analytics time after rewriting the job orchestrator
go + kafka + postgres
00 / INIT
I ship systems that stay quiet
Distributed auth, streaming pipelines, and backend services that don't wake you up at 3AM.
$ whoami
👋 I'm Sounish Nath
$ role --watch
Backend Engineer
Software Engineer II at Microsoft, previously Oracle and TCS. I design and ship distributed backend systems — auth infra, real-time streaming, and fault-tolerant microservices. Rewrote a job orchestrator in Go + Kafka −64% SQL time, moved 40M+ async tasks/day, cut infra cost −68%. Based in Bengaluru.
hosts
│
01 / TELEMETRY
Numbers from production, not a pitch deck
The systems I ship are measured in latency, cost, and whether someone gets paged at 3AM.
LIVE │ process=microsoft.fabric.spark│ role=SWE II│ loc=bengaluru
infra.cost
−68%
Infrastructure spend on 300GB/day batch + stream
pubsub · spark · delta
throughput
40M+
Async tasks moved off a brittle PHP service
per day, production
users.dau
1M+
Daily users on OCI streaming microservices
k8s · circuit breakers
- 01Microsoft· Fabric Spark infra──
- 02Oracle· auth, streaming, RCA sidecar──
- 03TCS· orchestration + data lake
02 / SELECTED SYSTEMS
Things I built because existing tools were boring
Consensus, caches, and real-time collab — the kind of work that belongs in a Git history, not a slideshow.
gossip-raft.go
Gossip Raft - KV Store
01Built a sharded, fault-tolerant key-value store from scratch in Go — implements Raft consensus for leader election and log replication, gRPC for inter-node RPC, and a gossip protocol for membership and eventual consistency. Designed to survive node failures without data loss.
GogRPCRaftLeader election
shortener.go
URL Shortner Service
02Designed a read-heavy URL shortener in Go — uses Redis for sub-millisecond lookup caching, PostgreSQL for persistence, and Bloom filters to eliminate DB hits for non-existent keys. Built to handle high read throughput with minimal write amplification.
GoPostgresqlRedisBloom Filters
multiplayer.ts
Leetcode Multiplayer
03This project is a multiplayer platform for practicing Leetcode problems. It allows users to collaborate and compete in solving coding challenges in real-time with integrated audio calling and cloud code execution.
GoHTMXWebRTCWebSockets
03 / STACK DUMP
What I actually type into a terminal
Languages, systems, and the boring-but-vital glue that keeps clusters honest.
~/sounish/stack.toml
utf-8 · 4 tabs
1 [programming_languages]
C#PythonGolangJavaTypeScriptSQL
2 [backend_and_systems]
ASP DOTNETNest.JSGo/EchoFastAPIDistributed SystemsMicroservicesREST APIgRPCApache KafkaApache Spark
3 [data_and_ai]
RedisPostgreSQLBigQueryLLMsVector DBsAI Agentic
4 [cloud_and_infra]
DockerKubernetesOCIGCPTerraformCI/CDPrometheusGrafana
5 [deeplearning_and_ai]
TensorFlowPyTorchKerasscikit-learnHugging Face TransformersKubeflowTorchserveLLMsGenerative AIlangchainCrewAI
04 / CODE REVIEW
What people wrote after shipping with me
Not testimonials. LinkedIn reviews from people who sat in the same incident channel.
pull/endorsements · 4 comments● approved
GenpactI have worked with Sounish within the same team (Analytics team) where he had joined as a fresher. But the work ethic and competency he had shown both on the technical front as well as on the functional front was quite remarkable and matched the level of any pro developer.
Prashant Mishra · Lead Data Engineer, Genpact · via LinkedIn
InfosysSounish Nath consistently demonstrated his skills through problem-solving, teamwork and actively participation. He played a pivotal role in Machine Learning development and showcased his ability to deliver expected outcomes. What I appreciate most about Sounish is his dedication, reliability, creativity about the
Arokia jenefer Antony · Senior Consultant of Infosys
Tata Consultancy ServicesSounish is hardworking and dedicated associate. He joined TCS and within no time print his footprint on our project. He is expert in Cloud Services almost in all areas with AI / ML expertise.
Udhav Lahane · Azure Data Engineer of TCS
DeloitteSounish is hardworking employee. He is always keen to learn new technologies and apply that knowledge in building new and innovative solutions. He is having good hands on knowledge about Google Cloud and has expertise in AI & ML fields. He is a great team player, mentor and leader who can be an asset to any
Rinkesh Kore · Data Engineer of Delotte USI
05 / OPEN A CHANNEL
If the problem is interesting, I will answer
Distributed systems, data platforms, weird latency graphs. Skip the small talk.
