Work

Projects I’ve Built

Real n8n automations, RAG systems, and AI tools — each solving a concrete problem and built to actually run in production.

Featured Project

📡

Trendy Content Intelligence Pipeline

A fully automated research-to-inbox system that fires every morning at 9am. It reads topic keywords from a Google Sheet, fires Tavily web searches for each topic with 24-hour recency filtering, passes 5 articles per topic to GPT-4o-mini for virality scoring and content generation (strict 100-word limit, 8-word hook), logs the output back to Sheets, then sends a styled HTML digest email — all without touching a keyboard.

Schedule (9am)→ Google Sheets→ Loop per topic→ Tavily Search→ GPT-4o-mini Agent→ Sheets Log→ Gmail Digest

n8n Tavily API GPT-4o-mini Google Sheets Gmail Schedule Trigger Virality Scoring

How it works

Schedule fires at 9am dailyReads all topic rows from the Topic_sheet in Google Sheets

Loop per topicBatches each topic through the same pipeline independently

Tavily advanced searchQueries last 24hrs of news — 5 results, 3 chunks each, no images

GPT-4o-mini virality agentScores all 5 articles, picks the most viral, writes an 8-word hook + <100-word post

Logged to SheetsTopic, date, and generated content appended to the logs tab

Gmail HTML digestAll topics aggregated and emailed as a styled branded report

🎧

Customer Support AI Agent

Monitors a Gmail inbox every minute. A text classifier (OpenRouter) routes emails — support queries go to a Gemini 2.0 Flash agent with a Pinecone-backed FAQ knowledge base; everything else hits a no-op. The agent auto-replies with grounded, policy-accurate answers. Zero human triage required.

Gmail Trigger→ Text Classifier→ Gemini Agent→ Pinecone RAG→ Auto-reply

n8n Gmail Gemini 2.0 Flash Pinecone Text Classifier RAG

📬

Gmail AI Triage & Sheet Lookup

Watches a Gmail inbox in real time, parses each incoming email snippet with GPT-4.1-mini, then uses a Google Sheets tool to look up relevant data and drafts a context-aware reply. The whole loop — receive, retrieve, respond — runs without any human in the loop.

Gmail Trigger→ GPT-4.1-mini Agent→ Sheets Lookup→ Auto-reply

n8n Gmail API GPT-4.1-mini Google Sheets AI Agent

🧠

RAG Pipeline + Knowledge Chatbot

Two pipelines in one. The ingestion side watches a Google Drive FAQ folder — when a new file lands, it downloads, chunks, embeds via OpenAI, and upserts into Pinecone under a dedicated namespace. The chat side uses Claude 3.5 Sonnet (via OpenRouter) with Pinecone as a retrieval tool to answer questions strictly from the knowledge base.

Drive Trigger→ Chunk + Embed→ Pinecone Insert

Chat Trigger→ Claude 3.5 Agent→ Pinecone Retrieve→ Answer

n8n Google Drive Pinecone OpenAI Embeddings Claude 3.5 Sonnet RAG

💬

↗

Custom GPTs

Two purpose-built GPT assistants published on the OpenAI GPT Store — each with a custom system prompt, curated knowledge base, and a specific audience in mind.

🔐 AI Security PhD Research Mentor

Guides researchers through AI security literature, methodologies, and phishing/social engineering defence.

🏗️ Prompt Architect

Helps engineers and researchers design, test, and refine high-quality prompts for production LLM pipelines.

OpenAI GPTs Prompt Engineering Knowledge Files GPT Store