Projects (IA)
Fenya
Fenya és una eina per buscar feina que permet pujar i gestionar diversos CVs, amb el càrrec i la ubicació. Pots editar-los, duplicar-los o esborrar-los per adaptar-los a cada oferta.
How I used AI
It is currently in the testing phase, if you want an access please write me
For each CV, Fenya adds offers from various sources (Adzuna, Jooble, JSearch/LinkedIn, Feina Activa and others) and sorts them by fit using semantic matching with LLMs.
The application (Next.js 16, Postgres/Drizzle) has been developed with the help of Claude Code as a coding agent, with a spec-driven flow and TDD.
Runtime Inference:
DeepSeek V4 Flash (via SDK compatible with OpenAI, with reasoning disabled and exit in JSON mode) extracts a structured profile from the CV, infers the search terms and scores each offer from 0 to 100 with a weighted rubric (skills 50%, experience 25%, location 15%, language 10%), justifying the note and indicating the skills you have and those you lack.
As a second evaluator, on demand, there is the Jev model (TypeSafe SDK). Jev does not generate free text: it chooses a level within a closed scale and indicates its trust.
In indexed rubric mode, it scores separately the same four criteria, each with five levels described.
In holistic mode, it makes a single global lace judgment. Thus the note is more deterministic and auditable, and you can compare three LLM "judges" on the same offers to see which one is closer to the human criterion.
- Claude Code
- Deepseek
- Jev
- Next.js
- Postgres