---
title: "Discussion on \"Why Every AI UI Crashes Mid-Stream (And the Schema-First Fix)\" | Hashnode"
description: "Discussion on \"Why Every AI UI Crashes Mid-Stream (And the Schema-First Fix)\". Streaming JSON from LLMs to UI Is Still  Broken. Here's How I Fixed It"
url: "https://hashnode.com/posts/streaming-json-from-llms-is-still-broken/69c37e6a8a263e79cbefc588"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Why Every AI UI Crashes Mid-Stream (And the Schema-First Fix)](https://vijay-singh.hashnode.dev/streaming-json-from-llms-is-still-broken?utm_source=hashnode&utm_medium=feed)

Streaming JSON from LLMs to UI Is Still Broken. Here's How I Fixed It

Mar 25 · 17 min read · vijay-singh.hashnode.dev  
Read the article

-   [#javascript](https://hashnode.com/tag/javascript)
-   [#llm](https://hashnode.com/tag/llm)
-   [#hlsstreaming](https://hashnode.com/tag/hlsstreaming)
-   [#ai](https://hashnode.com/tag/ai)
-   [#web-development](https://hashnode.com/tag/web-development)
-   [#reactjs](https://hashnode.com/tag/reactjs)

## Written by

[vijay singh](https://hashnode.com/@vjsingh)

@vjsingh · Rome , Italy

  
[View profile](https://hashnode.com/@vjsingh)

Staff Software Engineer & AI Architect | Bridging Applied AI and Product Engineering | AI Safety, Cost-Aware Systems & Architecture

I’m a backend-heavy Full Stack Staff Software Engineer and AI Architect, building production systems with React, TypeScript, and Node.js, alongside applied AI/ML platforms in Python. My foundation is backend engineering — designing distributed, observable, and cost-efficient systems that power real products at scale. While I work across the full stack, my strength lies in backend architecture, service design, data systems, and platform infrastructure. I specialize in translating AI capabilities into reliable production systems that hold up under real users, real data, and real unit economics. At Staff level, I operate across backend engineering, AI systems, and product architecture — bridging machine learning capabilities with robust software design. What I design and operate: • Scalable backend services and APIs (Node.js / TypeScript) • Full-stack product systems (React + TypeScript) • Production-grade generative AI applications (customer-facing products, internal tools, workflow automation) • LLM-powered data processing and enrichment pipelines • RAG systems grounded in structured business context • Knowledge graph integrations (Neo4j) for structured reasoning • Cost-aware inference systems (dynamic routing, semantic caching, usage optimization) • Secure and sandboxed AI execution environments • CI/CD and evaluation-driven ML deployment workflows I treat AI as infrastructure — not a feature. Safety, cost control, and context grounding are enforced at the systems layer rather than left to prompts. My impact is strongest in backend-heavy system design, platform thinking, and building long-lived technical foundations that teams can scale on. Core strengths: Backend Engineering • Full Stack Development • React • TypeScript • Node.js • Distributed Systems • AI Platform Engineering • Production ML & GenAI • RAG • Knowledge Graphs • System Architecture • Python

## Responses

Markdown supported · ⌘↵ to post

No responses yet.