Avani Agarwal

About

LLM/agentic systems practitioner who partners directly with customers to productionize prototypes end-to-end—defining business-aligned metrics, building evaluation datasets, and shipping monitored, cost-bounded inference APIs on AWS Lambda. Notably delivered a vehicle damage classification system that cut manual review by 40% and stabilized agent workflows by instrumenting state transitions to uncover and fix a race-condition-driven skipped tool call.

Experience

Education

Key Strengths

• Takes LLM prototypes to production with rigorous evaluation, monitoring, and cost controls
• Translates business risk into concrete model/system success metrics (precision by class, false negative rate, latency, cost)
• Improves model reliability via labeling standardization, structured outputs, and confidence-based fallbacks
• Optimizes production inference under AWS Lambda constraints (batching, INT8 quantization)
• Diagnoses complex LLM/agent failures with end-to-end tracing and targeted instrumentation (found race condition causing skipped tool calls)
• Builds guardrails and regression tests to prevent silent failures and recurrence
• Effective technical demos/workshops tailored to developer audiences; iterates based on feedback
• Partners with sales through live demos and technical Q&A to drive adoption and close deals
• Delivered measurable impact: reduced manual review by 40% in vehicle damage assessment workflow

Reference Highlights

• Highly recommended by reference
• Translates complex technical systems into reliable, production-ready solutions
• Strong customer-facing effectiveness; receives positive client feedback
• Efficient and organized
• Clear communicator with both developer and enterprise teams
• Able to bridge technical details to non-technical audiences
• Quickly assesses priorities
• Defines measurable success metrics
• Rapid iteration and execution under tight deadlines
• Balances technical depth with reliable solutions
• Strong ownership and cross-functional impact
• Drives performance and reliability improvements across teams
• Uses customer deployment feedback to shape product direction
• Delivered a production LLM solution for vehicle damage checking