Inspiration
I was inspired by how fragmented hiring workflows are: resumes live in one place, screening happens in another, and the browser is where most decisions are actually made. I wanted a single, browser-native copilot that helps candidates and recruiters move faster with clearer signal.
What it does
I learned a lot about combining LLM-based analysis with structured parsing, and how to design a system that works across a web app, API, and browser extension without feeling disjointed. One key takeaway was making outputs explainable and repeatable, not just "smart."
How we built it
I built the project as a FastAPI backend with an AI analysis pipeline, a Next.js frontend for onboarding and dashboards, and a Plasmo-based browser extension for in-context insights on job pages. The backend handles resume parsing, skills extraction, and code/job analysis. The frontend is a simple, fast UI. The extension is where the product feels alive. A rough mental model for matching looks like score=α⋅skills+β⋅experience+γ⋅signal.
Challenges we ran into
The hardest challenges were keeping data flow consistent across three surfaces, tuning AI output so it is useful without being noisy, and making the extension feel instant while still calling the API. Debugging cross-origin behavior and auth flows across the app and extension was also a big lift.
Accomplishments that we're proud of
Built a working end‑to‑end product: API, web app, and browser extension. Delivered real-time resume and job analysis where hiring actually happens. Designed a cohesive UX across three surfaces without losing clarity.
What we learned
The best AI results come from strict structure and clear prompts, not just bigger models. Cross‑origin auth and data flow are harder than the ML itself. Users trust insights more when the reasoning is visible and consistent.
What's next for supervoke
Complete the hiring fields to support full recruiter workflows. Expand beyond tech careers to multiple industries and role types. Prepare for launch: positioning, onboarding, and early access. Conduct market research to validate segments, pricing, and distribution.
Built With
- fastapi
- gemini
- httpx
- javascript
- next.js-14
- pdfplumber
- plasmo
- pydantic
- python
- radix-ui
- react
- react-18
- recharts
- tailwind-css
- tree-sitter
- typescript
- uvicorn
Log in or sign up for Devpost to join the conversation.