<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Sarthak V. Kumar</title><description>Thoughts and writings</description><link>https://sarthakvk.com/</link><item><title>Building Gokey A Key-Value Store on Raft</title><link>https://sarthakvk.com/writings/software/building-gokey-a-key-value-store-on-raft/</link><guid isPermaLink="true">https://sarthakvk.com/writings/software/building-gokey-a-key-value-store-on-raft/</guid><description>Building a replicated key-value store in Go on HashiCorp&apos;s Raft — how SET/DELETE turn into log entries, replicate across nodes, and apply via an FSM.</description></item><item><title>Building NSE tick collector with Zerodha Kite Connect</title><link>https://sarthakvk.com/writings/software/building-nse-tick-collector-with-zerodha-kite-connect/</link><guid isPermaLink="true">https://sarthakvk.com/writings/software/building-nse-tick-collector-with-zerodha-kite-connect/</guid><description>Building a real-time NSE tick collector with Zerodha Kite Connect — ~9,000 instruments over multiple websockets, batched to Parquet on S3, queried with DuckDB.</description></item><item><title>Designing a scalable fanout service</title><link>https://sarthakvk.com/writings/software/designing-a-scalable-fanout-service/</link><guid isPermaLink="true">https://sarthakvk.com/writings/software/designing-a-scalable-fanout-service/</guid><description>Designing a scalable fanout service that decouples event processing from delivery using Kafka, queues, and workers — linearly scalable to millions of recipients.</description></item><item><title>Scaling Acoustic Scene Simulation Pipeline</title><link>https://sarthakvk.com/writings/software/scaling-acoustic-scene-simulation-pipeline/</link><guid isPermaLink="true">https://sarthakvk.com/writings/software/scaling-acoustic-scene-simulation-pipeline/</guid><description>Building a distributed acoustic scene simulation pipeline for Audio ML — turning 10M scenes (~20 years of CPU and petabytes of audio) into a tractable workflow.</description></item></channel></rss>