• Python
• FastAPI
• JavaScript
• Chrome Extensions
• Gemini API
• Ollama
• RAG
• Multi-Agent Systems
• LLMs

Overview

RecruitSage Copilot answers the question students actually ask before a campus drive: is this worth applying to, and what will they ask me?

It lives inside the placement portal as a side panel. Open a company drive page and the extension extracts the role, CTC, eligibility, and skills, then runs a multi-agent research pipeline over campus history, public sources, and alumni networks.

Highlights

• 17-point red-flag audit — bonding clauses, offer revocation history, CTC inflation, stagnant tech stacks.
• Compensation breakdown — headline CTC versus realistic monthly in-hand take-home.
• Campus intel — previous visits, shortlist numbers, CGPA cutoffs, and the exact questions asked in past drives.
• Alumni links — one-click LinkedIn search for seniors who worked at the company.
• Local-first option — the same pipeline runs fully offline against a local Qwen model.

Architecture

Overview

A Python/FastAPI backend behind a JavaScript Chrome extension, with all research agents running in parallel and streaming results back into the side panel.

Agents

• Campus Intel — queries structured campus data for past drives, cutoffs, and question banks.
• Web Search — aggregates Glassdoor, AmbitionBox, Reddit, and news via a DuckDuckGo search agent.
• Alumni — resolves company → Thapar senior links for direct outreach.

Grounding

• RAG over 1,400+ campus interview questions, filtered by company and round, so generated prep material stays attached to real questions.
• Portal context injection into the system prompt for follow-up chat, reducing hallucination on role-specific details.

Model routing

• A hybrid router exposes one interface over Gemini API (cloud) and Ollama/Qwen (local, offline), streaming tokens in both cases.