⌁ SkillPulse
GitHub ↗

Candidate Match

See which software roles match your experience and where the strongest opportunities are.

Try a sample profile — See how SkillPulse matches different backgrounds to the market.
Backend Engineer 4+ years

Python, FastAPI, PostgreSQL, Docker, Redis, Celery & scalable REST APIs.

Junior AI Eng. 2 years

PyTorch, Hugging Face, pgvector, LangChain, RAG pipelines & embeddings.

Full-stack Eng. 3 years

React, TypeScript, Node.js, Next.js, PostgreSQL, Tailwind & modern state.

Cloud / DevOps 6+ years

Kubernetes, AWS, Terraform, Docker, Go, Linux systems & CI/CD infrastructure.

Jobs

2,481 roles

Search across structured market vacancies with lexical and dense semantic retrieval.

Try searching:
Active Vacancies
Loading vacancies...

Market

Real-time skill demand, seniority mix, and 30-day technology velocity across software markets.

Active Roles
2,481
Across Europe, UK & LATAM
Markets Tracked
7
Multi-regional telemetry
Normalized Skills
1,420+
Canonical ESCO & O*NET concepts
Top Skill Demand
Python (38%)
In backend & AI postings
Skills in Demand
Frequency across active vacancies
Loading technologies...
Emerging Demand
Fastest rising requirements
LLM & GenAI +31% velocity
Found in 31% of AI & Python vacancies
FastAPI +24% velocity
High concurrency backend standard
pgvector & RAG Rising
Leading semantic search requirement
Data updated recently · ESCO & O*NET taxonomy

How SkillPulse works

A transparent 5-step pipeline from multi-market job collection to explainable candidate matching.

01 — Collect

Ingestion

Jobs are ingested from multiple sources across European (Arbeitnow, Remotive, Himalayas, Jobicy) and Latin American markets.

02 — Extract

Extraction

Structured requirements are extracted through deterministic keyword tries and model-assisted extraction pipelines.

03 — Normalize

Taxonomy

Skills are mapped to canonical ESCO / O*NET concepts (e.g. reactjs → React, postgres → PostgreSQL).

04 — Retrieve

Retrieval

Lexical and semantic retrieval are combined using Reciprocal Rank Fusion ($k=60$) on PostgreSQL pgvector.

05 — Match

Matching

Candidate-job fit is calculated using explainable signals: 50% Hard Skills + 20% Role Alignment + 30% Dense Similarity.

Technical architecture & specifications → 
POSTGRESQL 16 + PGVECTOR
├── Lexical Indexing: English & Portuguese Full-Text Search (tsvector)
├── Dense Vector Search: 384-dimensional Embeddings (HNSW Index, Cosine Distance)
└── Reciprocal Rank Fusion: RRF(d) = 1 / (60 + rank_lexical) + 1 / (60 + rank_vector)

AI & EXTRACTION STACK
├── Tier 1: Aho-Corasick Deterministic Keyword Trie (<1ms)
├── Tier 2: Contextual Token Boundary NER Pattern Matcher (~5ms)
├── Tier 3: Cloud AI Router (Groq Llama 3.3 70B & OpenRouter) with structured JSON schemas
└── Normalization: Canonical ESCO / O*NET Skill Taxonomies

SkillPulse is built on FastAPI, PostgreSQL 16 with pgvector, Cloudflare Workers AI (BGE embeddings), and Groq / OpenRouter LLM endpoints.