Case study 02 · Personal · SyntaxilitY · AI platform
Tygent — Multi-Agent Recruitment AI
An AI recruitment platform where eight specialized agents analyze resumes and job descriptions, match candidates, flag hiring risks and assess interview readiness — streamed live to the recruiter as each agent runs.
8
orchestrated AI agents
The problem
Recruiting means reading resumes, understanding job descriptions, judging fit, spotting hiring risks and preparing interviews — work that is fragmented across tools and heavy on manual effort.
Architecture
- 018-agent orchestration pipeline, one agent per stage of candidate evaluation
- 02Real-time execution streaming to the browser over server-sent events
- 03Resilient LLM calls with model fallback, plus persistent analysis history
- 04JWT auth, Zod validation and PostgreSQL via Drizzle ORM
Recruiter dashboardNext.js · React
APINode.js · JWT · Zod
Live execution streamServer-sent events
Agent orchestrator8 specialized agents
Resume & JD agentsParsing · analysis
Matching & riskFit · hiring risk
Readiness & reportsRecommendations
LLM providerOpenAI · model fallback
PostgreSQLDrizzle · analysis history
Step through how a request moves through this system.
What I owned
- 01Architected the platform and the agent workflow
- 02Designed backend APIs and the data architecture
- 03Implemented resume / JD processing, candidate matching and risk-analysis flows
- 04Implemented SSE-based real-time execution streaming
- 05Implemented authentication, validation, persistence and execution history
- 06Integrated resilient OpenAI workflows with model fallback
Outcome
One AI workflow covering resume analysis, JD analysis, candidate matching, hiring-risk assessment, interview readiness, recommendations, analytics and reporting.
Stack
- Next.js
- React
- TypeScript
- Node.js
- PostgreSQL
- Drizzle
- OpenAI
- AI Agents
- SSE
- JWT
- Zod
- Docker