ARH Jarvis | Voice and Text Command Automation Assistant (n8n + Windows Agent) A personal automation assistant that takes voice or text commands from a web page or phone and carries them out on a real Windows computer, in English or Roman Urdu. Voice commands are transcribed through Groq, then converted into a structured command. Text commands go through a separate authenticated route: common commands are resolved with deterministic rules with no AI cost involved, while complex natural-language requests are planned by DeepSeek before execution. Both paths converge on the same Windows agent, connected through a permanent tunnel, with a confirmation step before any sensitive action runs. Current capabilities include opening and searching applications such as YouTube, Gmail, and WhatsApp Web, sending Gmail messages and WhatsApp messages, checking the PC's live status, writing files to approved folders, and monitoring the automation platform itself: it can detect the latest failed workflow execution, summarise the error, and save a report to the desktop. Every request is authenticated at each hop, and errors return as structured categories such as authentication_error, planner_error, or execution_error, rather than a plain crash. Actions such as arbitrary software installs, unrestricted system commands, or unrestricted file deletion are intentionally blocked by design, not because they were forgotten. This is a personal proof of concept still being hardened, with voice recognition consistency and fully autonomous multi-step flows as the current development focus. It uses the same architecture principles I bring to client systems: hybrid AI plus deterministic logic, authenticated multi-hop communication, and structured error handling. Stack: n8n (self-hosted on Contabo), Groq, DeepSeek, Windows local agent, ngrok, Gmail API.

0

37

AI Sports Stats Automation | n8n Data Pipeline + Parent Dashboard An automated data pipeline built in n8n that pulls sports statistics on a schedule, cleans and structures the data, and pushes it into a live dashboard for parents to track, with no manual data entry required. The workflow handles scheduled triggers, API pagination, data transformation, and structured writes to the dashboard's data source, with retry logic so one failed request does not break the entire run. Demonstrates the data engineering side of automation work: scheduled triggers, pagination, data cleaning, and dashboard-ready output, not only chatbots and lead generation. Stack: n8n, REST APIs, Google Sheets and dashboard integration.

0

69

Lead Generation & Contact Research Automation | n8n + Claude AI Agent Built a 90+ node workflow in n8n that takes a plain-language lead request from Slack and turns it into a verified, CRM-ready contact list, with no manual research involved. Claude AI parses the request into a structured ideal customer profile covering company size, industry, region, and role. The workflow then searches for matching companies and contacts, verifies each email through Hunter and MX record checks, and cross-checks speciality, location, and name relevance before anything is written to the sheet. Every record passes through multi-layer deduplication against the current batch, past results, and related company records, then is routed into Ready, Review, and Rejected tabs with a full audit trail and failure log. Result from a real run: the client asked for 50 qualified leads. The system delivered exactly 50 Ready contacts, held back 7 for manual review, and filtered out 143 low-confidence or duplicate records automatically. Stack: n8n (self-hosted), Claude AI, Serper, Hunter, Google Sheets, Slack.

0

117

WhatsApp AI Chatbot & Order Automation System | n8n + Claude AI Agent An n8n-based WhatsApp chatbot that handles customer conversations and order automation from start to finish, delivered as part of a completed 80-hour engagement rated 5.0 out of 5. Client feedback: "Very strategic and great to work with! Knows his stuff and communicates clearly." The agent uses Claude to understand customer intent, hold context across a conversation, and take the correct action, capturing orders, answering frequently asked questions, and updating records, while every interaction is logged for review. Built with the same reliability standard used across all of my automation work: input validation, structured AI output instead of raw free text, and error handling that catches failures instead of letting messages disappear silently. Stack: n8n, Claude AI, WhatsApp, webhooks and API integration.

0

110

Abdul Rehman Haris

AI Automation Engineer | n8n, APIs & AI Agents