Building with AI on your own systems: MCP servers, retrieval-augmented chatbots, semantic and product search, agents, and the scripts that automate search work.
A workflow for drafting alt text with a vision-capable model and approving it by hand: what good alt text looks like, a drafting script, the review step and privacy considerations.
Ashruba Pankhade ·
How to build natural-language product search that stays accurate: a model turns requests into validated filters, your catalogue supplies real results, and catalogue quality decides how well it works.
Ashruba Pankhade ·
How retrieval-augmented generation and fine-tuning differ, when each fits, the order to try them in, and the common mistakes that make business AI answers outdated or wrong.
Ashruba Pankhade ·
How to build a WhatsApp chatbot that answers from your own website content: the WhatsApp Business Platform, a working webhook, grounded answers, human handover, consent and data handling.
Ashruba Pankhade ·
A Python script that sends buyer questions to Claude and Perplexity through their APIs with web search, records brand and competitor mentions and cited URLs, and saves the results for trend tracking.
Ashruba Pankhade ·
How to replace keyword-only site search with hybrid semantic search in PostgreSQL, combine rankings with reciprocal rank fusion, and use visitors' searches to find content gaps.
Ashruba Pankhade ·
Six design principles for AI agents that work with your website's data: narrow tools, read before write, human confirmation, defence against prompt injection, limits and logging, and data protection.
Ashruba Pankhade ·
The architecture of a website chatbot that answers from your own pages: chunking, embeddings in PostgreSQL with pgvector, grounded answers with citations, safeguards and evaluation.
Ashruba Pankhade ·
How to expose your Laravel application to AI assistants through the Model Context Protocol with the official laravel/mcp package, and the security design that matters most.
Ashruba Pankhade ·