Manish Reddy

About

Software engineer (Yellow.ai) who built and productionized an AI-driven resume tailoring system using embeddings + Chroma RAG + QLoRA fine-tuning, deployed via Docker/Kubernetes with CI/CD on a CPU-only Oracle VM. Demonstrates strong reliability/evaluation rigor (custom hallucination/coverage/relevance metrics) and measurable business impact, including a 60% user satisfaction lift from improving chatbot intent accuracy with product and support teams.

Key Strengths

• Built and deployed an AI-driven resume tailoring system using RAG + vector DB + fine-tuned LLM
• Improved system accuracy by 45% using weighted evaluation (hallucination rate, skill coverage, relevance) on unseen datasets
• Production deployment under constraints (CPU-only Oracle VM) via quantized/CPU-optimized models and manual network rule configuration
• Strong MLOps/DevOps execution: Dockerized services, CI/CD with large-model artifact handling via object storage, optimized build times with caching
• Kubernetes rollout reliability practices (health-based traffic distribution, restart policies, rollback to last healthy image, near-zero downtime)
• Structured approach to AI agent reliability: stateless stages, unit tests per step, timeout/failure testing, replaying real user flows
• Cross-functional impact: partnered with product/support at Yellow.ai to improve chatbot intent accuracy and raised user satisfaction score by 60% in a quarter

Experience

Education

Skills

Languages