Shivam Shahi

Software Engineer

New Delhi, India

Product-minded Software Engineer focused on building scalable, high-performance applications and robust system architectures.

About Me

I engineer high-performance software systems with a strong emphasis on scalable architectures and robust backend infrastructure. My expertise spans hybrid-cloud deployments, real-time event-driven systems, and building intuitive, seamless user interfaces.

Operating effectively in high-velocity, early-stage environments, I take ownership of the full engineering lifecycle—from architecting complex data pipelines and orchestrating microservices to shipping polished, production-ready applications.

When I am not deep in code, you can find me playing soccer, exploring new music, or sharing technical insights within the developer community.

Experience

A brief about my work experiences and professional journey.

•As the first engineering hire, I drove the technical vision and foundational architecture from inception to production.
•Architected a hybrid-cloud platform using a custom Go-based edge agent, seamlessly integrating cloud infrastructure with legacy on-premise ERPs like Tally.
•Engineered and automated end-to-end workflows for contract execution and document state synchronization via Slack and DocuSign APIs.
•Designed and implemented robust real-time SSE infrastructure, ensuring instant state synchronization for invoices, payments, and reconciliations.
•Developed scalable multi-stage ingestion pipelines for invoice processing, automated approvals, and complex journal entry reconciliations.

ReactNext.jsTypeScriptTailwind CSSNode.jsGo

Projects

Some of the projects I've built and deployed.

ShadowLLM

Production AI reverse proxy for zero-latency traffic bifurcation. Safely shadow-tests experimental prompts and cheaper models against live production traffic with side-effect virtualization and real-time semantic drift scoring.

TypeScriptFastifyRedis StreamsPostgreSQL (Neon)+8

Payment Processing Engine

High-throughput concurrent payment orchestration engine in Java 17 LTS. Features atomic multi-key idempotency locks, fine-grained thread synchronization, dynamic multi-currency fee calculation strategies, state machine transition guards, and resilient exponential backoff retry policies.

Java 17Low-Level Design (LLD)SOLID PrinciplesConcurrency & Thread Pools+6

Concurrent Task Scheduling Engine

Thread-safe asynchronous task execution engine with Directed Acyclic Graph (DAG) cycle detection dead-lock prevention. Implements deterministic priority queues with atomic sequence tie-breaking, bounded queue admission control, worker thread fault isolation, and graceful zero-leak shutdown.

Java 17MultithreadingDAG Cycle Detection (DFS/BFS)Priority Blocking Queues+5

Unified Multi-Cloud Python SDK

Multi-cloud serverless developer SDK and CLI providing a single unified API abstraction over disparate serverless runtimes (Cloudflare Workers, AWS Lambda). Features AST code serialization, automated packaging, local emulation runtime with FastAPI, and extensible provider adapter architecture.

Python 3.11SDK ArchitectureAdapter PatternAST Code Serialization+5

Realtime Collaborative Kanban Board

A Kanban Board Application built with React, Redis, and WebSockets for real-time task management and multi-client board synchronization.

ReactJavaScriptSupabase (Postgres)Upstash (Redis)+6

Ai Symphony

A platform providing AI tools like image generation, summarization, and chat using open-source models deployed on edge workers.

TypeScriptTailwind CSSNext.jsCloudflare Workers+4

Bug Issue Tracker

A comprehensive bug tracking system with role-based authentication, project management, and real-time issue tracking.

ReactNode.jsExpress.jsSQLite+5

Prototype YC landing page for Preloop

Engineered a high-performance prototype marketing page for Preloop, a YC W24 backed startup, focusing on modern UI/UX principles and animations.

ReactNext.jsMongoDBTailwind CSS+4

Writing

All articles

Deep dives into distributed systems, backend architectures, and real-world system design.

HashnodeSeptember 29, 202616 min read

How YouTube Processes a Video After You Upload It

From a 2 GB upload to transcoding, queues, adaptive bitrate streaming, and CDN delivery.

“You open YouTube Studio, pick a 2 GB file, and click Upload. The bar fills. A few minutes later the video plays in more than one quality, on a phone, a laptop, and a television. That looks like one operation. It is not. Between the click and the first frame someone else can stream, the platform has to accept a large file over a bad network, store it, encode several versions, build metadata, survive machine failures, and serve the result from an edge CDN near the viewer.”

System DesignDistributed QueuesAdaptive Bitrate (HLS/DASH)TranscodingEdge CDN

Technical Arsenal

I view languages and frameworks simply as tools to solve complex business problems. While I adapt quickly to any ecosystem, I have deep production expertise in a few core areas:

Frontend & Product: React, Next.js, TypeScript, and modern state management. I build interfaces that feel instantaneous and native.
Backend & Systems: Node.js, Python, and Go. I design scalable microservices, real-time event-driven architectures, and secure API gateways.
Data & Infra: PostgreSQL, Redis, Docker, and hybrid-cloud deployments. I focus on high availability, efficient querying, and resilient deployments.