About

I combine generative AI logic modeling with user experience design best practices to create agile, effective, and humanized conversational interactions. Managing end-to-end user journeys, intent validation mapping, and deterministic prompt guardrail architectures under strict international criteria (ISO/IEC 42001).

Services

Core functional domains and conversational interaction specialties.

01 // INTERACTION
Conversational flows

Designing fluid end-to-end multi-channel user paths using UX best practices and intent logic structures to optimize engagement.

02 // ALIGNMENT
Prompt engineering

Architecting structural Markdown-based prompt guardrails and behavioral guidelines to anchor LLM responses and protect data integrity.

03 // RESEARCH
UX research & diagnostics

Facilitating co-creation workshops, mapping personas, conducting usability diagnostics, and tracking conversational metrics.

Experience

Professional history across technical content design, user interaction systems, and stakeholder facilitation.

2026 — Present

Conversational content designer & AI implementation consultant

Askmanu · Bucharest, Romania
2022 — Present

Technical writer & UX contributor

Oracle · Bucharest, Romania
2021

Proposal writer & non-key expert (UX research lead)

ACZ Consulting · Bucharest, Romania
2019 — 2021

Academic documentation editor (process optimization)

MDPI Sustainability · Cluj, Romania

Portfolio

Empirical, process-driven case studies mapped across key technical proficiencies.

1. Conversational AI architecture

Conversational logic & prompt safety architecture

Platform: Askmanu Copilot Interface · Source: Interaction logic deployment

Challenges
  • User intent deviations causing model anomalies.
  • Dynamic outputs lacking structured prompt constraints.
  • Interaction loops failing under ambiguous phrasing inputs.
Solution
  • End-to-end multi-channel conversational path design.
  • Deterministic text pattern verification rule logic.
  • Markdown-based prompt guardrails ensuring model context validation.
Recommendations
  • Map core user personas directly within the intent validation layer.
  • Isolate diagnostic outputs to avoid conversational drift errors.
  • Enforce rigorous interaction metrics to secure flow visibility.
Process
Identify tracking parameters to build robust interaction flows. Architect baseline configurations and conversational logic layers inside specialized system settings, keeping operational oversight of prompt validation to achieve 95% response accuracy and traceability.

2. UX & conversation design

Interactive workspace navigation redesign

Platform: Oracle Clinical One Hub · Source: UX Architecture Proposal

Challenges
  • Dense navigation layouts caused immediate cognitive overload for software operators. 
  • The absence of context-aware progressive disclosure forced unnecessary manual validation across nested layout frames.
Solution
  • Integrated collapsible context menus and logical layout divisions to display active configuration fields sequentially while filtering secondary fields from the main view canvas.
Recommendations
  • Utilize progressive disclosure models as a baseline system style. 
  • Keep visual focus locked onto active task rows to accommodate natural attention limits and minimize search lag.
Process
  • Mapped end-to-end interaction paths and updated system screens iteratively in Figma. 
  • Presented layout shifts directly to technical squads.
  • This workspace design initiative contributed to being awarded the 2024 Indigo Design Gold Winner for UX, interface, and navigation.

Prototype: Relational context ingestion engine (second brain)

Platform: Custom Claude Environment · Source: Independent AI Orchestration Project

Challenges
  • Unstructured text variables and personal data logs were too dense for basic search indexing, causing context window fragmentation across active knowledge tasks.
Solution
  • Simplified extensive source inputs into structured data modules, leveraging system-level profiling parameters to direct the model's focus toward immediate milestones.
Recommendations
  • Prioritize cross-referencing metadata structures over raw wording parameters. Establish clear relational tagging rules alongside domain filters to avoid model retrieval errors.
Process
  • Audited complex context parameters with prompt diagnostic tests. Engineered optimized self-service workflows that reduced processing lag by 20%. Validated structural prompt logic, demonstrating the practical scalability of modern LLM context windows.

Tone of voice architecture & persona curation

Platform: Askmanu Conversational Copilot · Reference: personality slider.png

Challenges
  • Automated error notifications trigger immediate user friction. 
  • Verbose system replies or unclear context explanations degrade user efficiency under strict development timeline constraints.
Solution
  • Designed a direct interaction matrix aligned with corporate criteria, stripping out polite filler text to ensure conversational system outputs remain short, actionable, and verifiable.
Recommendations
  • Address system warnings with immediate corrective steps. 
  • Enforce rigorous ISO 26000 corporate accountability guidelines to secure structural workflow visibility and brand alignment.
Process
  • Audited system conversation parameters and integrated clear Markdown-based guidelines across chat prompts, achieving 95% response accuracy and traceability.
Specialist


Language agnostic
Pragmatic developer


Edgy / quirky
Rigid / tough


Supportive / patient
Directive / authority


Peer / collaborative
Authority


Approachable

3. UX research & user diagnostics

Cross-border user journey mapping & discovery

Platform: European Union Institutional Framework · Source: User discovery project

Challenges
  • Isolate core workflow friction points toward regional platform adoption while structuring complex user feedback data across 24 separate target user groups.
Solution
  • Constructed interactive persona models and user journey maps, establishing a balanced development roadmap across multi-stakeholder requirements.
Recommendations
  • Analyze qualitative research data concurrently. Remove complex jargon loops when presenting diagnostic findings to central decision-making panels.
Process
  • Categorized user feedback parameters on shared digital whiteboards. Led extensive qualitative UX research tracks with over 100+ institutional stakeholders, creating blueprints via Axure to shape product choices impacting 12.4 million citizens.

National administrative system modernization

Platform: ANFP Institutional Capacity Initiative · Source: UX Diagnostics summary

Challenges
  • Frontline service personnel workflows were slowed by fragmented interaction models, uncentralized operational guidelines, and system usability blocks.
Solution
  • Formed a high-volume data ingestion architecture to process, validate, and index sensitive qualitative user feedback databases delivered by departments nationwide.
Recommendations
  • Run direct validation diagnostics with field end-users to align administrative system interaction layouts with international benchmarks, directly impacting a target population of 19+ million citizens.
Process
  • Conducted country-wide focus groups and stakeholder sessions. Engineered optimized workflow wireframes and interface scripts via Balsamiq, using usability tests to verify fast field adoption.

Achievements

Formal awards and continuous system validation milestones.

2024

Gold winner: UX, interface & navigation

Indigo Design Awards · Workspace initiative contribution
2024

Special recognition: Inclusive design excellence

Oracle · Accessibility design implementation
2017

Innovation award: Best user-centric content

Bucharest · Consumer platform taxonomy design

Contact

Available for full-time Conversational Design roles, UX content architecture positions, and product team alignment contracts.

Email connection eimarin@icloud.com
Professional network linkedin.com/in/elizamarin