Software Engineering student building AI-powered automation — agents, retrieval systems, and self-hosted workflows that run on their own, on real schedules, touching real tools.
Second-year Software Engineering student at Air University, Multan, spending most evenings building AI automation instead of scrolling.
Most of what I build lives on a self-hosted VPS — n8n, Postgres with pgvector, and Ollama running open models — wired into Gmail, LinkedIn, Calendar, and Drive through agents that actually get things done, not just chat about them.
A Boss–Sub Agent system: one orchestrating agent delegates to specialized sub-agents for calendar, contacts, and email/file handling, each with its own memory — extended with a RAG pipeline so it can answer questions straight from stored documents.
Finds a current AI/automation topic through web search, drafts a LinkedIn post with an AI agent, and pauses for my approval by email before anything goes out. Runs itself every Monday morning.
Pulls open roles from 16 companies across five ATS platforms, scores each one against a resume with AI, and surfaces only the strongest matches — no blind auto-apply, just a shortlist worth reading.
A self-hosted resume screening workflow running entirely on Ollama, deployed via Docker on a custom domain — no candidate data ever leaves the server.
A clean Streamlit app for translating text between languages, built with an automatic fallback translator so it keeps working even when the primary service briefly isn't.
An end-to-end lead qualification pipeline in the works: incoming leads get validated, deduplicated, and scored by an AI reasoning step, then routed automatically down one of several paths — voice follow-up, human review, nurture, or archive — with a full audit trail at every step.
Open to freelance work, internships, and anyone who wants to build an agent that actually does the job.