Meet Ustaad اُستاد
اُستاد — Your AI-Powered Career OS
The career coach Pakistan never had.
Ustaad is an AI-powered career intelligence platform built for tech graduates across Pakistan. Upload your CV, pick your target role and city, and Ustaad's 7-agent pipeline delivers a real-time market snapshot, honest skill gap analysis, a personalized learning roadmap, an ATS-optimized CV rewrite, and a scored mock interview — all in under 5 minutes.
Scroll
The Problem
Why We Built This
Every year, thousands of CS and IT graduates in Pakistan enter the job market without a clear picture of what employers actually want. Generic CVs get filtered out by ATS before any human reads them. Skills don't match market demand. Interview prep is guesswork. And there's no single tool that understands the Pakistani tech job market deeply enough to give real, actionable guidance.
Ustaad changes that. We built it to be the career partner that every Pakistani tech graduate deserves — honest, intelligent, and grounded in live market data from Rozee.pk, Mustakbil, and LinkedIn Pakistan.
7
Specialized AI Agents
~5 min
Full Pipeline Runtime
100%
Pakistan-Grounded Data
User Journey
How It Works
Four simple steps from CV upload to a downloadable career dossier.
Upload Your CV
Drop your PDF resume and tell us your target role and city. That's all the input you need.
Pipeline Runs Live
Watch 7 AI agents process your data in real-time via Server-Sent Events — no page reloads.
Get Your Analysis
Review your Readiness Score, skill gaps, ATS-optimized CV, and 30/60/90-day learning plan.
Practice & Export
Complete an AI mock interview, then download your full PDF Career Dossier.
Architecture
7 AI Agents. One Career Dossier.
Each agent is a specialized Gemini 2.5 Flash instance — sequential pipeline with parallel execution on Agents A4 + A5 for maximum speed.
Profile Agent
Parses your PDF CV into a structured JSON profile — education, experience, skills, and projects.
Market Intelligence Agent
Runs 4 live Serper.dev queries on Pakistani job boards (Rozee.pk, Mustakbil, LinkedIn) for real-time demand data.
Gap Analysis Agent
Compares your profile against market demand and computes a quantified Readiness Score (0–100).
Roadmap Agent
Builds a milestone-driven 30/60/90-day learning plan with free resource links to close skill gaps.
CV Rewrite Agent
Rewrites your CV bullet points and summary for ATS optimization with a before/after comparison view.
Interview Agent
Runs a multi-turn mock interview with 15 tailored questions and scores each answer on a 1–10 rubric.
Export Agent
Aggregates all pipeline outputs into a styled, downloadable PDF Career Dossier using ReportLab.
Deliverables
What You Get
Six concrete outputs — all generated from a single CV upload.
Readiness Score
0–100 readiness benchmark vs. live market data
Gap Analysis Report
Critical vs. nice-to-have skills with priority rankings
30/60/90 Roadmap
Milestone-driven learning plan with free resource links
ATS-Optimized CV
Rewritten CV with before/after comparison view
Mock Interview
15 tailored Q&As with per-answer scoring and feedback
PDF Career Dossier
Full ReportLab-generated export of all findings
Kaggle × Google
Course Concepts Demonstrated
Ustaad is a direct application of 4 core agentic paradigms from the Google AI Agents Intensive course.
Multi-Agent Collaboration
7 specialized agents in a sequential + parallel dependency graph. Agents A4 and A5 run concurrently via asyncio.gather(), cutting pipeline time by ~40%.
asyncio.gather(run_roadmap_agent(), run_cv_rewrite_agent())Tool Use & Real-Time Grounding
The Market Agent uses Serper.dev to run 4 live Google Search queries per session against Pakistani job boards — grounding the pipeline in current market reality.
"Python Developer jobs Karachi site:rozee.pk"Structured Outputs & Schema Hardening
Every agent enforces strict JSON output via Gemini's response_mime_type API. A 4-layer fallback parser (json → fence strip → regex → ast.literal_eval) prevents crashes.
response_mime_type="application/json"Production-Grade Async Pipeline
Live SSE streaming via asyncio.Queue. Exponential backoff with 8 retries. Automatic failover to a secondary Kaggle-hosted Qwen model if Gemini quota is exhausted.
wait = min(2 ** attempt, 15) # backoffStack
Built With
Submitted to
AI Agents: Intensive Vibe Coding Capstone 2026
Track: Agents for Business / Agents for Good
Creators
The Team
Built by two developers from Karachi, Pakistan 🇵🇰
Syed Muhammad Hassan
Backend Engineer & AI Agents
GenAI Developer and Co-Founder of AgenticEra Systems. Built the entire 7-agent pipeline, FastAPI backend, orchestration logic, and the Kaggle Qwen model failover system.
A product by AgenticEra Systems — AI automation for the real world.