Dhanush Chinivar

"Small business owners drown in spreadsheets, Notion pages, and inboxes then still can't answer 'which product made me the most money last quarter?' AI BI Agent changes that. Ask a plain English question, and a four-agent LangGraph pipeline (Data → Analysis → Insight → Action) pulls live data from Google Sheets, Notion, and Gmail via MCP, reasons across all of it, and delivers a real answer not a dashboard you have to decode yourself. Every Monday at 8am, n8n fires an automated report before you've had your coffee."

"E-commerce teams drown in thousands of reviews and still can't answer 'what are customers actually complaining about this week?' ReviewIQ turns that noise into signal. It ingests reviews from Shopify and CSV/Excel through an event-driven serverless pipeline (API Gateway → SQS → Lambda → DynamoDB), then uses Amazon Bedrock (Claude) to auto-generate weekly reports on sentiment, recurring themes, severity, and anomalies. Built multi-tenant from day one with Clerk auth, KMS-encrypted tokens, and least-privilege IAM, all defined as infrastructure-as-code with AWS SAM."

"Imagine dumping your entire company knowledge base into a chatbot, then realising any employee can now ask it to summarise the CFO's salary review or pull a confidential security audit. Role-Aware RAG Chatbot kills that risk it semantically routes every query to the right knowledge domain (finance, HR, engineering, marketing), then hard-checks the user's role permissions before a single token is returned. Right data for the right role. A clean 403 for everyone else."

"Tracking stocks manually, missing price swings, and reading generic market news that has nothing to do with your portfolio. Rallify fixes that a real-time stock platform with live candlestick dashboards, custom price-alert thresholds, and an AI Daily Digest that hits your inbox every morning with a personalised market summary built from your own watchlist."

"Most RAG pipelines treat images as an afterthought convert them to captions, lose half the detail, and call it multimodal. This system skips the middleman entirely. By embedding both PDF text and images into CLIP's shared vector space, a plain text question can retrieve the exact chart, diagram, or figure it's asking about no captions, no information loss. Claude then reasons over the retrieved text and visuals together to produce answers that actually reflect what's on the page."

"Imagine tracking your macros manually every week, writing grocery lists by hand, and still blowing your budget by Wednesday. MealsForge eliminates that chaos an AI that designs your personalised weekly plan, auto-builds your shopping list aisle-by-aisle, and keeps your streak alive with adherence analytics and nutritional nudges, all behind a frictionless subscription."