Sabita Kumari

About

Built and deployed JobMatcher AI, an LLM-driven workflow automation product for job seekers that extracts requirements from job descriptions, matches to user skills, and generates tailored outreach. Demonstrated strong production engineering by cutting per-run cost ~70%, improving reliability with retries/backoff/fallbacks, and reducing hallucinations via schema validation and templating; also orchestrated the system with LangGraph plus Docker Compose across API, vector DB, and workers.

Experience

Education

Key Strengths

• Built and deployed LLM-powered workflow automation tool (JobMatcher AI) to production
• Reduced per-workflow effort from ~25 minutes to under 2 minutes
• Cut LLM run cost ~70% by routing tasks to lightweight models, batching, and reusing context
• Improved system stability via retries, exponential backoff, structured logging, and provider fallback
• Reduced hallucinations using schema validation and predefined templates
• Designed state-based orchestration with LangGraph across ingestion, extraction, matching, generation, and automation steps
• Debugged and stabilized multi-container deployments (API, vector DB, workers) using health checks, startup dependencies, and centralized env config
• Metrics-driven evaluation; identified extraction failures via task success rate and improved reliability with hybrid parsing + validation
• Effective collaboration with non-technical stakeholder using demos/check-ins to translate user pain points into requirements