4 views

TalentMatch AI – LLM-Powered Recruitment Platform

AI-powered resume screening and semantic job matching platform

TalentMatch AI is an AI engineering platform for resume screening and job matching. It combines NLP, LLMs, embeddings, semantic search, vector databases, and explainable AI to analyze candidate profiles, evaluate job requirements, calculate match scores, identify skill gaps, and provide candidate matching insights.

PythonFlaskReactTypeScriptLangChainOpenAIAnthropicHugging FaceSentence TransformersPostgreSQLChromaDBRedisCeleryDockerGitHub Actions

Code

Problem

Recruiters need to evaluate large numbers of resumes against job requirements, but traditional keyword-based screening can miss semantic relationships between candidate experience, skills, projects, and job requirements.

Solution

I designed an AI-assisted talent matching platform that combines resume parsing, skill extraction, embeddings, vector search, semantic matching, LLM-based analysis, and explainable scoring to evaluate candidate-job compatibility.

What I Built

AI-powered resume parsing
Candidate profile and skill extraction
Job description analysis
Semantic job matching
Resume and job embeddings
Vector search with ChromaDB
LLM-assisted candidate analysis
Candidate ranking
Explainable match scoring
Skill gap analysis
Resume optimization
AI evaluation workflow
REST API with Flask
React and TypeScript frontend
Asynchronous processing with Celery and Redis
Docker-based development environment
CI workflow with GitHub Actions

Architecture

┌──────────────────────┐
                         │    React Frontend    │
                         │   TypeScript/Vite    │
                         └──────────┬───────────┘
                                    │
                                    ▼
                         ┌──────────────────────┐
                         │     Flask REST API   │
                         │      Python          │
                         └──────────┬───────────┘
                                    │
                  ┌─────────────────┼─────────────────┐
                  │                 │                 │
                  ▼                 ▼                 ▼
          ┌──────────────┐  ┌──────────────┐  ┌──────────────┐
          │ Resume       │  │ Job          │  │ Matching     │
          │ Parser       │  │ Analyzer     │  │ Engine       │
          └──────┬───────┘  └──────┬───────┘  └──────┬───────┘
                 │                 │                 │
                 └─────────────────┼─────────────────┘
                                   ▼
                         ┌──────────────────────┐
                         │ Embedding Pipeline   │
                         │ Sentence Transformers│
                         │ / Hugging Face       │
                         └──────────┬───────────┘
                                    ▼
                         ┌──────────────────────┐
                         │ Semantic Search      │
                         │ ChromaDB             │
                         └──────────┬───────────┘
                                    │
                     ┌──────────────┴──────────────┐
                     ▼                             ▼
             ┌──────────────┐              ┌──────────────┐
             │ LLM Layer    │              │ PostgreSQL   │
             │ OpenAI /     │              │ Application  │
             │ Anthropic    │              │ Data         │
             └──────┬───────┘              └──────────────┘
                    │
                    ▼
             ┌──────────────┐
             │ Explanation  │
             │ & Ranking    │
             └──────────────┘

       Redis + Celery → Asynchronous AI Processing

Capabilities

Full Stack Engineering

ReactTypeScriptViteTailwind CSSFlaskREST APIJWTTanStack Query

Backend & Data

PythonFlaskSQLAlchemyPostgreSQLChromaDBRedisCeleryVector SearchDatabase Design

AI Engineering

LLM IntegrationNLPEmbeddingsSemantic SearchRAG-related RetrievalLangChainOpenAIAnthropicHugging FaceSentence TransformersExplainable AIAI EvaluationCandidate Matching

Delivery & Quality

DockerDocker ComposeGitHub ActionsAPI TestingCIAsynchronous Processing

Engineering Decisions

Problem	Decision	Reason
Keyword matching can miss semantic relationships	Implemented embedding-based semantic matching	Allows candidates and jobs to be compared based on meaning rather than exact keywords
Candidate and job information requires semantic retrieval	Used ChromaDB for vector search	Enables efficient similarity-based retrieval
Different LLM providers may be required	Designed the AI layer to support OpenAI and Anthropic	Provides flexibility in model selection and provider integration
AI processing can be computationally expensive	Used Redis and Celery for asynchronous processing	Prevents long-running AI operations from blocking API requests
Candidate matching needs explainability	Implemented weighted matching and explanation generation	Helps users understand why a candidate matches a position
Resume processing requires structured candidate information	Implemented resume parsing and skill extraction	Converts unstructured resume content into information usable by the matching engine
AI-assisted recruitment requires responsible use	Designed the system as decision support rather than autonomous hiring	Keeps human decision-makers involved and avoids using protected attributes

My Contribution

Designed the overall AI application architecture
Designed the resume-to-job matching workflow
Implemented the Flask REST API backend
Implemented AI and NLP processing workflows
Integrated LLM providers including OpenAI and Anthropic
Implemented embedding generation and semantic search
Integrated Sentence Transformers and Hugging Face
Implemented vector search using ChromaDB
Designed the candidate matching and scoring engine
Implemented candidate ranking and skill gap analysis
Designed the PostgreSQL data model
Implemented asynchronous processing using Celery and Redis
Developed the React/TypeScript frontend integration
Containerized the application using Docker and Docker Compose
Configured CI workflows using GitHub Actions
Documented system architecture, API specifications, database schema, and AI pipeline

Results

MVP DevelopmentSemantic MatchingVector SearchLLM-powered AnalysisAI-assisted Candidate RankingAutomated Skill Gap Analysis