The Output - 1
The Output - 2
The Frontend UI
VS code SS

Inspiration

Finding relevant internships, hackathons, and scholarships often means wasting hours wading through expired postings, complex prerequisites, and scattered application deadlines. We wanted to build a tool that eliminates this friction for students and fresh graduates. By automating the search, eligibility validation, and ranking process, we aimed to transform chaotic online listings into clear, prioritized career opportunities.

What it does

Taskmaster is an autonomous multi-agent AI system designed to streamline opportunity discovery. Users simply input their search criteria (e.g., software engineering internships or hackathons open to postgraduates) along with their student identifier. The platform then deploys a pipeline of specialized agents that gather, evaluate, and format matching listings—delivering a clean, ranked output complete with bold deadlines, target criteria, and direct application links.

How we built it

Challenges we ran into

API Rate Limiting: Handling multi-agent API calls in sequence triggered 429 RESOURCE_EXHAUSTED rate limits during initial testing. We resolved this by optimizing our agent model selection and implementing strategic pauses (time.sleep) in the orchestrator pipeline. Agent Context Handoff: Ensuring that raw output from the Researcher Agent was cleanly parsed by the Eligibility Checker without loss of detail required refining system prompts for each specialized agent. Git Security & Secret Scanning: Accidental staging of local environment files (.env) triggered GitHub's Secret Scanning Push Protection. We resolved this by configuring .gitignore rules and rewriting clean commit histories.

Accomplishments that we're proud of

End-to-End Automation: Successfully created a fully functional, 3-agent autonomous workflow that runs in seconds. Sleek UX: Designed an intuitive, responsive Streamlit dashboard featuring live agent execution cards and styled outputs. Robust Multi-Agent Architecture: Built a modular system where each agent has a clear, specialized role, making the system easy to maintain and scale.

What we learned

Agentic Design Patterns: Learned how to decompose complex tasks into focused sub-tasks assigned to distinct LLM agents rather than relying on a single monolith prompt. Quota & Latency Management: Discovered techniques for managing free-tier API rate limits in sequential multi-step workflows. Clean Code Practices: Gained hands-on experience in securely managing secrets, configuring repository rules, and structuring multi-file Python projects.

What's next for Internship & hackathon AI Finder

Personalized Student Profiles: Enable users to upload resumes so agents can dynamically tailor eligibility checking against individual skill sets. Automated Deadline Notifications: Integrate email or Discord notifications to alert students before tracked application deadlines close. Expanded Data Sources: Incorporate web scraping APIs to pull directly from job boards, Devpost API, and university career portals in real time.