Edosa Leta
Summary
Senior UX Designer specializing in conversational AI and generative UI for enterprise products. Combines a CS/NLP background with prompt engineering and design systems governance to build the skill frameworks and component contracts that keep AI-generated interfaces accessible, on-brand, and structurally reliable at scale.
Experience
Senior User Experience Designer · IC3
Oracle · Redwood City, CA- Defined Oracle Redwood’s AI interaction pattern library, establishing guidelines for response formatting and trust standards adopted by product teams.
- Designed a versioned skill-governance system for enterprise applications using an eight-intent taxonomy and automated lint gates, ensuring accessible UI.
- Created user flows and interaction patterns for multi-step AI agents, collaborating with engineering for precise implementation.
- Developed an AI accessibility agent based on Oracle’s Redwood guide, leading research on bias across demographics to inform inclusive design.
- Published agent evaluation guidelines addressing task completion, response quality, and failure modes, giving teams effective quality measurement frameworks.
- Contributed interaction patterns and component guidelines for chat and Oracle AI Agent Studio, integrating them into the Redwood design system for broader reuse.
User Experience Designer · Intern
Oracle · Redwood City, CA- Conducted 12 one-hour user research sessions using Oracle’s digital assistant, uncovering insights into conversational interaction patterns to inform design improvements.
- Proposed and designed a research concept for a digital assistant that includes diverse English-speaking user groups, ensuring broader user accessibility.
- Researched unconscious bias in conversational AI to identify gaps and propose mitigations for bias in digital assistant models, enhancing model fairness.
Projects
English–Amharic neural machine translation
Capstone project · African Leadership University- Trained and iterated an English–Amharic neural machine translation model using JoeyNMT across 20+ training runs (about 6 hours each).
- Collected and translated about 10,000 entries from a source dataset, improving translation accuracy by 5% on the BLEU score.
Education
M.S. in Computer Science
University of Maryland, Baltimore County · 2023
Coursework: Human-Centered Computing, Design Thinking, Design and Analysis of Algorithms, Advanced Computer Networks
B.S. in Computer Science
African Leadership University, Kigali · 2021
Skills
AI interaction designConversational AI governanceGenerative AI prompt engineeringUser research methodologiesProduct managementUsability testingDesign patterns
Tools: Figma, Jira, Oracle AI Agent Studio, Codex