Work
About
Contact
Resume
Edwin Fernandez
Edwin Fernandez
Work
About
Contact
Resume
Product Design · Enterprise · AI/ML
Improving
the
Log Analyzer
Redesigning Panacea.ai — an AI-powered log analysis platform for Nutanix Site Reliability Engineers
to unify fragmented workflows and cut root cause analysis time by 87.5%.
COMPANY
Nutanix
DURATION
8 Months
ROLE
Product Designer
TOOLS
Figma · FigJam
PHASE
Phase 1 · Shipped
SCROLL TO EXPLORE
87.5%
Reduction in root cause analysis time
45
UAT users in pilot testing
405
Log bundles analyzed in testing
8min
Average processing time per
bundle
01 — THE PROBLEM
Fragmented tools.
Delayed diagnoses.
When a customer opens a support case, SREs had to juggle NCC, Panacea,
Insights, Remote Commands, Logs, and Service pages simultaneously — none of
which spoke to each other. Every minute spent context-switching was a minute a
customer waited.
sre_workflow_before.log — The painful reality
[09:14:02] ALERT: Customer VM down — cluster xyz-prod-01
[09:14:05] → Open Panacea.ai ... scanning log bundle
[09:22:18] → Switch to Insights ... correlating raw metrics
[09:35:44] → Open NCC ... cross-checking health checks
[09:48:01] → Remote Commands ... pulling CVM service status
[10:03:17] → Back to Panacea ... re-reading log signatures
[10:14:52] RCA still incomplete. 1hr+ elapsed.
// AFTER REDESIGN
[09:14:02] ALERT: Customer VM down — cluster xyz-prod-01
[09:14:05] → Panacea.ai Upload bundle → AI Summary generated
[09:22:05] ✓ Root cause identified. RCA drafted. 8 minutes.
No Unified View
Log data, cluster health, CVM metrics, and historical patterns existed in silos. There was no single pane of glass where an SRE could see the full picture of a cluster incident at once.
Junior SRE Gap
Senior engineers could intuit correlations across tools. Junior SREs could not — leading to inconsistent troubleshooting quality, longer case resolution times, and excessive escalations.
02 — THE RESEARCH & DISCOVERY
2 months of
structured learning.
Designing for an unfamiliar domain required deep immersion. I embedded
myself in the product, the team, and the workflow before drawing a single frame.
STAKEHOLDER 01
SRE Team
Real-time cluster monitoring without tab-hopping
Quick incident resolution with AI-guided triag
Consistent process regardless of experience level
Script Wizard for common cluster operations
STAKEHOLDER 02
PM
User experience as a priority metric
Rapid feature prioritization and delivery
Clear visibility into design decisions
Roadmap alignment across all teams
STAKEHOLDER 03
Panacea Dev Team
Richer AI-powered summaries and conclusions
Strict adherence to design system guidelines
Detailed API contracts per feature
Technical documentation for each component
"Panacea AI provides a great foundation for building an enterprise-level troubleshooting tool for SREs. The tool is able to quickly parse large amounts of log data, identify known issues, and provide a view into key events. The RCA summary is a first step into providing an automated Root Cause Analysis tool."
— Pilot SRE Tester, Nutanix UAT Program
OOUX Exercise with Senior SRE
Object-oriented UX sessions helped map out every data entity in the system — bundles, signatures, CVM nodes, alerts, layers — and how they related to each other. This became the backbone of the information architecture.
PRD & Grooming Sessions
Deep dives into the Product Requirements Document alongside recurring PM grooming sessions ensured design decisions were anchored in real requirements — not assumptions.
Glean AI for Domain Knowledge
Used Glean AI to rapidly build understanding of Nutanix-specific infrastructure concepts — AOS, hypervisors, CVM services, NCC — enabling me to speak the same language as the SREs I was designing for.
Deep Exploration of Existing Tools
Hands-on use of both Panacea.ai and Insights revealed the specific moments where engineers lost context, made errors, or gave up and escalated. These friction points became our design targets.
03 — DESIGN PROCESS
How might we unify
without
overwhelming?
The HMW question guided every design decision: create a platform that serves
diverse technical skill levels, blends AI capabilities naturally, and stays scalable without
a fixed end-vision.
PHASE 01
Discovery
2 months
Stakeholder interviews, PRD analysis, domain learning, OOUX exercise
PHASE 02
Define
10 days
Report architecture, data mapping, requirements with Senior SRE
Phase 03
Ideate
10 days
FigJam flows, sketches, first-draft
UI explorations
Phase 04
Design
3 months
High-fidelity UI, iterative sessions with SREs, design system alignment
Phase 05
Test & Ship
~2 months
45 UAT testers, 405 bundles, feedback loops, iteration
04 — KEY DESIGN DECISIONS
What we built,
and why it matters.
Every feature was born from a real pain point uncovered during research. Each
addresses a specific failure mode in the original workflow.
NEW FEATURE
Panacea Reports
A completely new module enabling SREs to create and combine log bundles for cross-bundle correlation.
Turns isolated snapshots into a continuous cluster health narrative — critical for
multi-incident pattern detection.
AI-POWERED
AI RCA Summary
The most critical deliverable. Natural-language summaries of anomalies, ranked by severity,
with affected CVM IPs highlighted. Transforms raw log data into
an actionable diagnosis in seconds — not hours.
VISUALIZATION
Heatmap & Event Timeline
A temporal visualization of log anomaly density and cluster events.
Lets SREs immediately identify when an issue started and what preceded it
replacing manual log scanning with pattern recognition at a glance.
INTEGRATION
Raw cluster metrics and CVM configuration data from Insights are surfaced directly within Panacea
— eliminating the #1 context switch that previously
broke SRE focus during triage.
CONVERSATIONAL AI
Ask AI
Raw cluster metrics and CVM configuration data from Insights are surfaced directly within Panacea
— eliminating the #1 context switch that previously
broke SRE focus during triage.
AI FEEDBACK
Add Rule Mechanism
Contextual AI chat enabling SREs to query log data in plain language and surface related KB articles.
Reduces the gap between junior and senior SRE troubleshooting
capability significantly.
05 — IMPACT & RESULT
From 1 hour
to 8 minutes.
87.5%
Reduction in root cause
analysis time
Simplified log analysis via automatic signature detection — eliminating manual pattern matching
Reduced need for manual log investigationthrough AI-curated
log views
Accelerated case triage time with unified bundle + Insights
view
Standardized troubleshooting quality across all SRE experience levels
Multi-bundle analysis enabled previously impossible cross-incident correlation
ClickHouse migration delivered performance and cost savings across engineering
45
UAT Testers
405
Bundles Analyzed
55
Combo
Bundles
8min
Avg Process Time
CUMMUNICATION CHALLENGE SOLVED
Prepared detailed design documentation for all pages, functional specs aligned with Panacea dev team API contracts, and update decks for PM to communicate changes to wider audiences — resolving the three-way communication breakdown between design, engineering, and product.
05 — LEARNINGS
What this project
taught me.
Every feature was born from a real pain point uncovered during research.
Each addresses a specific failure mode in the original workflow.
01
Documentation as Design
Writing clear, accessible design documentation wasn't overhead — it was the primary communication channel that unified three teams with conflicting priorities and vocabularies. Every page had a living spec.
02
Consistency Over Intensity
Recurring, methodical sessions with the end user (Senior SRE) produced more reliable design outputs than intensive sporadic sprints. Cadence created trust. Trust created candor. Candor created better design.
03
Use the Product You Design For
Extensively using Panacea.ai and Insights as a quasi-user revealed UX issues that no brief or PRD would have surfaced. Lived experience in the product generated the most valuable design hypotheses.
04
Multi-Perspective Reviews
Presenting designs within the design team before SRE reviews caught assumptions early and diversified solutions. Peer critique reduced the number of revision cycles with engineering significantly.
05
Design for Scalability First
With no fixed end-vision, every component had to be extensible. This meant advocating for a design system approach and creating modular patterns that could accommodate AI capability growth over time.
06
Design-Driven Feature Discovery
Script Wizard, CVM Configuration, and Insights data integration were all discovered through design exercises — not the PRD. Structured exploration with stakeholders is a legitimate product discovery method.