ATS Pipeline Architecture & Demo

Industrial-Grade Custom ATS Pipeline utilizing NVIDIA NIM and ChromaDB.

Pipeline Architecture Mapping

This diagram maps out every technical step, micro-decision, and database event needed to build an in-house, enterprise-ready resume screening system. It relies entirely on powerful Node.js libraries, mathematical cosine similarity matching, and NVIDIA’s free tier open-weight models (NVIDIA NIM).

In-Memory Vector MatchingNVIDIA NIM (NV-Embed-QA)Node.js (compromise / pdf-parse)Natural NLP (TF-IDF)NVIDIA Llama-3 Insight Engine

1. Ingestion & NLP Parsing

Uses Node.js pdf-parse for high-accuracy raw layout text extractions. It feeds text chunks straight to the compromise NLP library to clean up and extract named entities, explicit dates, and structural resume segments.

2. Vector Embeddings

Chunks are sent via standard API requests to NVIDIA NIM calling the open-weight nvidia/nv-embedqa-e5-v5 model. Resulting high-dimensional vectors are evaluated live in-memory via pure mathematical cosine distance logic.

3. Hybrid Scoring Blend

Calculates ranking using three metrics: 50% Dense Vector semantic matching, 30% Natural.js TF-IDF exact keyword mapping matrix overlays, and 20% deterministic rule-based heuristic checks.

Internal Architecture: Llama 3.1 8B Structured JSON Feedback Engine

This diagram exposes the internal structures of the most advanced component in the pipeline: the LLM-driven ATS Evaluator. It demonstrates how raw resume context, mathematical match scores, and strict JSON schemas are injected into the NVIDIA NIM Llama-3.1 8B model to generate the dynamic, section-wise breakdown grid seen on the frontend.

Comprehensive Process Diagrams

Explore detailed models including State Machines, Entity-Relationships, and Component architecture.