Sushma Sri B

About

Built and shipped production systems spanning real-time operational dashboards and an LLM-powered internal documentation assistant using RAG (embeddings + vector DB). Demonstrates strong focus on reliability and iteration: implemented guardrails and evaluation loops (human review, hallucination tracking, regression prevention) and improved performance/scalability through query optimization, caching, and retrieval tuning.

Key Strengths

• Owned features end-to-end from user feedback to launch and iteration
• Pragmatic tradeoff decisions (polling first vs full streaming) to ship quickly while managing operational complexity
• Data-driven iteration using usage metrics to identify API load/latency issues
• Performance and scalability improvements via query optimization, caching, and API tuning
• Production RAG system design (ingestion → embeddings → vector search → grounded generation)
• Strong LLM reliability/guardrails (grounding, confidence filtering, fallbacks, monitoring/logging)
• Built evaluation loops with real user queries, human review, and regression tracking
• Improved LLM answer quality by tuning retrieval (chunking and similarity thresholds)
• Designed scalable pipelines for messy real-world data with reliability and observability (async processing, retries, monitoring)

Experience

Education

Skills

Languages