Generative AI with LangChain: What I'd Ship Before Agents
Generative AI with LangChain connects models, retrieval and tools; here's how I'd choose the smallest useful workflow and test it before production.
Tech blogReact · AI · Architecture238 articles
Tech blog · India
Tech blogs on code, architecture and AI.
Deep dives into React.js, AI and ML integration, MERN stack patterns, and lessons from building production applications at scale in India — written by Deepak Kumar.
What it covers
- javascript92 articles
- node.js83 articles
- Engineering75 articles
- devops56 articles
+ 474 more topics in the archive
238 articlesLatest · Nov 2026
Showing 23 articles tagged AI/ML. Clear filter
The AI training pipeline separates knowledge from assistant behavior; I explain the stages and how I would use that distinction in production decisions.
Read the articleArtCraft's AI-Built Adobe Alternatives: What I'd Trust Today
ArtCraft offers seven open-source Adobe alternatives built with AI agents; I'd borrow its engineering process before trusting the apps for professional work.
What Is AGI? Why I'd Still Build for Uneven AI Today
Artificial general intelligence remains hypothetical; I explain what separates it from today's AI and how I'd evaluate capabilities before expanding autonomy.
RAG vs CAG vs KAG: What I'd Choose for Production AI
RAG retrieves documents, CAG reuses cached context, and KAG follows connected facts. I compare their production trade-offs and the checks I'd run before choosing.
Gemini 4 Carbon: What I'd Test Before Trusting Its Fixes
Gemini 4 Carbon targets vulnerability discovery and patching; I'd evaluate its evidence, permissions and total review cost before trusting it with production code.
LEANN Vector Index: Smaller RAG Storage, Slower Retrieval
LEANN cuts RAG index storage by recomputing embeddings during search. I look at the latency cost, build requirements, and checks I'd run before adopting it.
AI Coding Tools Need Smaller Tasks, Not Less Code Review
AI coding tools can shift effort from writing to verification. Here is how I would scope their tasks, review their output, and decide where not to use them.
Rust inference infrastructure: where I'd draw the line
Rust inference infrastructure targets serving, tokenization and data movement, not a wholesale Python replacement. I'd start with one measured bottleneck.
AI Can't Reverse SHA-256: Bitcoin Stays Safe in 2026
Despite claims, AI models can't reverse SHA-256 hashes to steal Bitcoin. They mimic lookup tables, not breakthroughs. Quantum risks differ.
NVIDIA Nemotron: Building AI Models Your Way
NVIDIA's Nemotron family offers a toolkit for creating specialized AI models. Explore how Nemotron changes AI deployment and model customization.
Pluckdown: Transform Pages into AI-Ready Markdown
Pluckdown converts pages, PDFs, and threads into clean Markdown for AI models. No account needed. Learn how it helps your AI workflows.
About the author
Deepak Kumar
Sr Software Engineer at India Today Group | Aaj Tak
Building scalable web applications and Generative AI systems. Writing about React, Node.js, AI and the engineering lessons from 9+ years of production work.
More about meHiring, or need something built?
Open to senior full-stack and AI engineering roles, and to focused contract work. Delhi NCR or fully remote. I reply within 24 hours.
What I take on
Full-stack product build
React, Next.js, Node, NestJS, MongoDB and MySQL — from idea to production.
AI features that ship
OpenAI and LangChain, RAG, semantic search, content and voice automation.
Architecture and scale
Microservices, real-time pipelines and caching — proven at national news scale.
Founding engineer work
Idea to MVP to scale, with SEO and analytics built in from the start.