Ruben Cespedes
Menu
TL;DR
Problem: Storage admins were spending too many manual steps across disconnected tools to identify and reclaim unused storage, with no single place to see, approve, and act on recommendations. What I did: Led end-to-end design of the Reclaimable Storage capability for Synapse, Dell's new agentic ITOps platform, while also helping shape its foundational navigation, chat interface, and UI patterns alongside a cross-functional team. Outcome: Shipped a functional MVP in under 4 months, demoed live by Dell Vice Chairman Jeff Clarke on the DTW 2026 main stage to strong customer engagement and excitement.
Role & Scope
As Lead Product Designer for the Synapse Reclaimable Storage initiative at Dell Technologies, I worked within a cross-functional team of 4+ designers, product managers, engineers, and stakeholders over approximately 4 months to deliver an MVP for DTW 2026. My deliverables spanned the full design process, from mind maps and user flows to low-fidelity wireframes, high-fidelity designs, and markdown spec files for engineering handoff.
Context & Problem
Synapse is Dell's next major step toward an autonomous, agentic ITOps platform, unifying Dell AIOps, MSM, and DAP to transform how enterprise storage admins observe, understand, act, and optimize across their full infrastructure stack.
Storage admins like Joe, the primary Synapse user, manage dozens of systems daily. Identifying unused storage objects that could be safely reclaimed required navigating multiple tools, running manual queries, and executing CLI or API commands that most admins preferred to avoid. Recommendations existed in the system but were buried, disconnected from action, and required too many steps to act on. The result: reclaimable storage sat unused, pushing organizations closer to capacity limits and unnecessary hardware spend.
Constraints
This work came with real limitations:
Hard 4-month timeline to deliver a demo-ready MVP for DTW 2026
Phase 1 scope limited to visibility, approval, and async execution, no new ML or agentic logic
Recommendation generation reused existing rule-based AIOps logic (no rebuild)
Actions gated by existing MSM/DAP authorization model, Synapse UI could not handle auth
Designed for a platform still being built, foundational UI patterns were being defined in parallel
Research
The team began with 12 proposed Synapse capabilities. Through collaborative sessions with PDMs, stakeholders, and engineers, we pressure-tested each one against three criteria: technical feasibility within the timeline, readiness of existing backend systems, and customer value at demo. That process narrowed the scope to 3 priority capabilities for DTW 2026. Reclaimable Storage was selected because the recommendation engine already existed, the execution path through MSM/DAP was proven, and the use case was immediately relatable to any storage admin in the room.
For Reclaimable Storage specifically, I mapped the full user journey, from the moment Joe receives an email or Teams notification about a recommendation, to logging into Synapse, reviewing the recommendation, approving the action, and tracking the async job to completion. That mapping exposed the key friction: too many disconnected handoffs between notification, context, and action.
Key Findings
Rough User Flow
Before defining any screens, I mapped how a user might encounter a reclaimable storage recommendation across multiple entry points, email, Teams, the Synapse home screen, a dashboard, or the assistant. This helped establish a shared mental model early and made clear the challenge was a system of interactions, not a single flow.
Defined User Flow
The rough flow was then broken down across actors and touchpoints, from a recommendation surfacing, through review and confirmation, to agent-driven execution and audit logging. This gave product, engineering, and platform teams a precise shared reference before moving into design.
Lo-fi Designs
Lo-fi screens were produced quickly to validate core interaction patterns without over-investing in polish. They covered the key states of the workflow: recommendation summary, assistant guidance, object selection, deletion progress, and post-action confirmation.
Approach
I started with mind maps to align the team on the full scope of the capability, entry points, states, edge cases, and decision logic. From there I moved into user flows covering every path Joe could take: email → Synapse, Teams notification → Synapse, direct login, and assistant-driven interaction via the Synapse chat interface.
Low-fidelity wireframes focused on the layout of the dynamic center panel, the heart of the Synapse experience, and how the Reclaimable Storage workflow would present recommendations, impact summaries, action buttons, and job status. I worked closely with engineers throughout, delivering annotated hi-fi designs and markdown spec files that mapped directly to the API and agent behavior specs.
In parallel, I contributed to foundational Synapse decisions alongside other designers and stakeholders: the left-rail navigation model, the chat interface interaction patterns, response output formats, and the overall UI framework that all capabilities would build on.
Final Designs
Solution
The Reclaimable Storage experience in Synapse gives Joe a single, intelligent surface to see every reclaimable storage recommendation, understand the impact, and act, all without leaving the platform.
When a recommendation exists, Synapse surfaces it proactively in the dynamic center panel with a human-readable summary: what can be reclaimed, which system it's on, how much storage is saved, and what percentage of used capacity that represents. Joe can approve a single object or select multiple using filters and act in bulk. Actions run asynchronously via MSM/DAP in the background, with job status tracked in the Tasks section and a summary sent to the full storage admin team on completion.
If a system isn't onboarded or authorized, the recommendation remains visible but the action is clearly disabled, with enough context for Joe to understand why and what to do next. Every action is logged for compliance and explainability. The Synapse assistant can also drive the entire workflow conversationally, Joe can ask, get a pre-built dashboard, approve, and track without touching a single menu.
Outcomes
Shipped a functional, demo-ready MVP in under 4 months alongside a cross-functional team of designers, engineers, and PDMs
Dell Vice Chairman Jeff Clarke demoed Synapse live on the DTW 2026 main stage, strong customer engagement and excitement from the audience
Customers were described as curious and enthusiastic about the platform's agentic direction
Synapse is now in active development with additional capabilities being added and iterated based on DTW feedback
Contributed foundational UI patterns, navigation, chat interface, response outputs, that now underpin the entire Synapse platform
Reflection
Designing for an agentic platform is fundamentally different from designing a feature inside a known product. The interface has to accommodate multiple entry points, probabilistic assistant behavior, and actions with real infrastructure consequences, all while staying approachable enough for a daily-use tool. The constraint of reusing existing backend logic in Phase 1 was actually a design advantage: it forced the focus onto the experience layer, where the most friction lived. The biggest lesson was how much foundational alignment work, on navigation, patterns, and response formats, is required before a single capability can feel coherent. That cross-team investment is invisible in the final demo but it's what made DTW possible.