Great design isn't just about what you make—it's about who you make it with.
I partner with those who value thoughtful, intentional work that speaks with purpose. If you're ready to be heard, I'm ready to listen.
Product designer & AI enthusiastic
Marco da silva portfolio
Hi, I'm Marco
I'm a product designer driven by curiosity — about people, interfaces, and how technology shapes everyday experiences.
I spend a lot of time observing how people use products, collecting interfaces I admire, and obsessing over small details that make things feel right.
Outside of work, you'll usually find me skateboarding, exploring visual ideas, or saving screenshots of things that inspire me.
Principal designer at
Wordsmith.ai
I was the founding designer at Wordsmith.ai, helping shape the product from the very beginning.
I focused on designing clear, well-crafted interfaces and interaction patterns to make complex systems easier to understand and trust.
I owned the core chat and document workflows and built the design system from scratch to support speed, consistency, and scale. I worked closely with the CEO to turn early ideas into a cohesive, shippable product.
Core strengths
Designing clear, usable interfaces that help people understand and use complex products with confidence.
High attention to visual detail and interaction quality, from layout and typography to motion and states.
Exploring ideas quickly through hands-on prototypes to learn, test, and move forward with clarity.
Building scalable systems that help teams move fast without sacrificing quality or consistency.
Working closely with founders and product teams to turn vision into practical, shippable experiences.
AI infor legal teams
Fashion e-commerce
Travel experiences
AI agents & workflows
Design systems
Marketplaces
Working together
You can definitely notice the effort and love in everything that Marco does. His attention to detail and visuals, making sure he's on top of the latests patterns and web trends, makes him a very important asset to the team.
What stood out most was how easy it was to collaborate with him. Marco brings a thoughtful, strategic approach to his work, and he's always coming up with innovative ideas that push projects forward. He takes real ownership of what he works on, communicates clearly, and consistently delivers.
He quickly became a highly trusted member of a newly formed mobile team and worked very closely with Product and Engineering to ensure that the TravelPerk app was not only useful for travellers but also well-designed. Very humble in nature, Marco could be trusted to deliver high-quality work without fuss.
Places where I worked in the last 6 years
I was the founding designer, working closely with the CEO and early team to shape the product from the ground up — from vision and UX to brand and core workflows.
- Shaped the core AI chat and document workflows
- Defined interaction patterns for complex systems
- Supported the product through a $24M funding round
Worked on behaviour and interaction design for a Slack-native product, focusing on how the system responds, guides users, and builds trust through conversation.
- Designed AI prompts and response patterns for managers in Slack
- Defined message structures, tone, and feedback loops
- Brought web UX thinking into conversational experiences
Helped define the product vision and design system for TravelPerk's native mobile apps, growing them from early concepts into a core product used at scale.
- Built iOS and Android apps from the ground up
- Designed and evolved a multi-platform design system
- Raised the bar for UI quality and interaction polish
Across these roles,
I've worked closely with product and engineering to shape design systems, bridge web and native experiences, and apply strong craft to AI-driven products.
At Letgo, I help the revenue model through cross-functional design.
At TravelPerk, I led the design of native mobile apps from the ground up as a Staff designer.
At Wordsmith and Oli, I designed AI-first products, defining interaction models and product foundations.
Native mobile at scale
I led the evolution of TravelPerk's mobile app into a scalable, fully native iOS and Android product, supporting critical travel workflows across teams and platforms.
Problem
TravelPerk's mobile app treated travel as a single, static experience.
In reality, travellers move through distinct states — planning, travelling, and post-trip — each with very different needs, urgency, and constraints.
As a result, critical information was often buried when it mattered most. Navigation reflected product features rather than user intent.
Mobile was useful — but not yet essential.
Challenge
Redesign mobile around real-world travel context — not screens or features — while maintaining quality and consistency at scale.
This wasn't just a navigation problem. The challenge was to support fundamentally different travel states without fragmenting the product or slowing teams down.
- Different travel states required different information density and urgency
- Time-critical moments demanded clarity over completeness
- The system needed to adapt contextually without breaking consistency
Designing for context, not screens
Solution Framing
We reframed the mobile app around travel context, not features.
Instead of a single, static experience, mobile adapts to where the traveller is in their journey — surfacing the right information at the right moment.
- Booking → explore and compare
- Travel → act quickly and stay informed
- Post-trip → wrap up and follow up
Mobile became a context-aware companion, not just a feature list.
Designing for context, not screens
Travel doesn't follow a linear flow. It shifts across moments, locations, and levels of urgency.
We redesigned mobile navigation from a fixed set of screens into a context-aware system that adapts to where the traveller is in their journey.
This allowed navigation, hierarchy, and information density to change dynamically — without breaking the overall product or mental model.
- Immediate access to critical trip information
- Contextual updates based on journey stage
- Clear, glanceable content under time pressure
- Proactive alerts and reminders
- Reliable access, even on the move
From static navigation to a state-aware system
Once context became the primary input, the information architecture had to change.
Instead of exposing the same navigation and hierarchy at all times, the system adapts based on the traveller's current state – planning, travelling, or post-trip.
State-aware navigation
Navigation priorities change based on where the traveller is in their journey.
Adaptive information density
Critical information surfaces automatically when time and attention are limited.
Single mental model
The product remains consistent while behaviour adapts underneath.
Contextual cards
Streamlined detail pages
Build travel experiences
Booking on the go
Hotels on the go
Support & chat
Goals
Increase usage when travellers are on the move — not just at booking time.
Enable check-in, tickets, directions, and support directly in context, reducing friction and reliance on web or support channels.
Support multiple travel states (planning, travelling, post-trip) without fragmenting navigation, ownership, or team velocity.
Impact
When users had an upcoming booking, the Home Hub was used on 90%+ of active days, making mobile the default entry point while travelling.
The "Upcoming" card alone accounted for 25-30% of total service views, despite being a single component.
By surfacing actions in context, usage increased across key moments: • Train tickets: ~50% → 55-60% • Flight check-in: ~20% → 25-26% • Directions: ~10% → 14-15%
Key learnings
Focusing on real user states reduced friction, clarified decisions, and made mobile feel essential in the moments that mattered most.
A single, intent-aligned entry point can drive disproportionate engagement.
Over several years, I led the evolution of TravelPerk's mobile app — rethinking how it looks, behaves, and supports travellers when it matters most.
2024 - Moving forward
This feature will introduce two ways users can interact with the results, a list view, and a map view.
This allows the user to easily browse through the hotels in the area without having to zoom in and out, drag the map, or manually select markers on the map. It also allows the user to quickly compare hotels in the same area and make a decision.
Goals (business & product)
What we were optimising for
- Increase confidence in hotel selection
- Reduce time spent evaluating and comparing options
- Shift more hotel searches and bookings to mobile
- Improve conversion without increasing cognitive load
Impact
Key outcomes
- Higher engagement with map view during hotel searches
- Clear preference shift toward map-based exploration
- Improved conversion for users engaging with location-aware search
- Strong qualitative feedback around clarity and ease of choice
You can see from the design, the company office is automatically setting the context of the search, and the user is getting a map first experience. As the user moves around the map, the hotel options also change, and visa-versa. This simplicity and elegance is what users expect.
Excerpt from a company blogpost
Hotels at TK
A focused redesign that helped users understand where a hotel is — not just what it costs.
Problem
The hotel search experience was primarily list-based and price-driven.
While functional, it failed to answer one of the most important questions users had when booking a hotel:
Is this hotel actually in the right place for my trip?
As a result, users spent more time cross-checking locations, switching views, or abandoning promising options due to uncertainty.
Users were comparing hotels without understanding location context.
Challenge
How might we help users quickly understand hotel location relative to what matters to them - without adding complexity to the booking flow?
- Serving both exploratory and decisive booking behaviours
- Supporting mobile and web consistently
- Scaling across cities, regions, and inventory sizes
Improve clarity without slowing down conversion.
Solution Framing
We reframed hotel search around spatial understanding, not just filters and lists.
Instead of treating maps as a secondary tool, we explored a map-first experience that:
- Anchors hotels to meaningful reference points (office, meeting location, city centre)
- Lets users visually compare options at a glance
- Reduces the mental effort of "imagining" distance
From list → map as an optional toggle To map → list as a complementary detail view
Scaling mobile design across teams and platforms
I led the design system effort in close partnership with engineering — defining patterns, tokens, and workflows that could scale with the product.
Design system at scale
TravelPerk's mobile apps scaled rapidly across iOS & Android. Teams were shipping fast, but UX consistency and velocity were at risk.
- Multiple verticals (flights, hotels, trains...)
- Two native platforms & Web app
- Multiple teams working in parallel
Design system
50+ components — Figma to Github handoff, iOS & Android.
—
Coming soon
Designing a global e-commerce experience at scale
From fragmented patterns to a coherent, multi-platform design system. At Farfetch, I worked on the redesign of the core e-commerce experience for a global luxury marketplace used by millions of customers worldwide.
Designing a global e-commerce experience at scale
From fragmented patterns to a coherent, multi-platform design system.
At Farfetch, I worked on the redesign of the core e-commerce experience for a global luxury marketplace used by millions of customers worldwide.
The outcome was a cohesive design system and e-commerce experience that aligned brand, product, and engineering—and became the foundation for future iterations across Farfetch's ecosystem.
It was a complex, long-running effort that strengthened my experience designing high-traffic, business-critical products at global scale.
letgo
I led the design of several verticals, from monetization to payments and shipping. I was responsible for leading a series of innovative initiatives aimed at monetization and transforming how design operates in letgo.
I help with the system
As the letgo team grew, it became increasingly important to maintain a consistent style and visual language across all areas of the product. With iOS, Android, and PWA it was clear that we needed more systematic ways to guide and leverage our collective efforts.
Designing AI-assisted contract drafting and review
From free-form chat to structured, trustworthy legal workflows.
My focus was turning AI outputs into something lawyers could confidently use.
Problem
Drafting legal documents is slow, high-stakes, and cognitively demanding.
Lawyers think in intent and outcomes — but ultimately work in documents.
Wordsmith already solved the starting point: lawyers could describe what they needed in natural language and get a strong, legally sound response.
Chat was powerful — but documents were still the source of truth.
Challenge
We needed to support the full lifecycle of legal work: drafting, reviewing, redlining, and exporting — without losing the benefits of chat.
This wasn't just a UI problem. We were bridging two mental models:
- Chat, where lawyers express intent in natural language
- Documents, where legal work is reviewed, negotiated, and trusted
Designing for documents — without losing the benefits of chat.
Outcome
Instead of treating chat and documents as separate experiences, we designed a single, continuous legal workflow — where intent flows naturally into a document, and the document stays connected to AI throughout its lifecycle.
- Chat captures intent and guides drafting
- A dedicated draft canvas becomes the source of truth
- Review and redlining happen directly in the document
- Final outputs are exported in trusted Word and PDF formats
AI became part of real legal work — not just the first draft.
Designing for documents, not chat
Legal work doesn't happen in isolated messages. It happens in documents — reviewed, negotiated, redlined, and ultimately sent outside the system.
Chat was the right entry point for intent, but documents had to become the source of truth. That meant rethinking the product around a persistent draft, not transient AI responses.
- Chat captures intent, documents carry trust
- Drafts persist across iteration, review, and negotiation
- AI stays present throughout the document lifecycle
- Outputs match the reality of Word and PDF workflows
Natural language first, clarification when it matters
The user starts by describing what they need in plain language. Before drafting, Wordsmith asks a small number of targeted questions to make sure the document is accurate, contextual, and legally appropriate.
AI generating the draft
Wordsmith generates the first draft, then steps back. The lawyer sees the emerging document and can refine the prompt or let the AI continue.
Document ready for review and iteration
From that point on, the document behaves like any other legal document — editable, reviewable, and easy to refine with or without AI.
Edit like a document. Improve like AI.
Once the draft is created, the document behaves like any other legal text. Users can format, restructure, and edit directly — or ask AI to improve specific sections without losing control of the content.
One document, multiple outputs
Lawyers can generate emails, cover letters, or supporting documents directly from the contract they're working on — without switching tools or copying content.
Related outputs
- Emails
- Cover letters
- Supporting documents
- ....
Review and redline with AI — clause by clause
Lawyers can review the draft directly in place, ask AI for clause-level improvements, and accept or reject changes individually — while keeping full control over the final document.
Goals
Ensure lawyers always end in a real, review-ready document — not a chat response — so AI supports how legal work is actually done.
Support drafting, iteration, redlining, and follow-ups in a single continuous workflow, reducing context switching between chat, Word, email, and review tools.
Enable AI to stay connected throughout the document lifecycle — drafting, review, and export — without breaking trust, control, or legal standards.
Impact
Lawyers could go from "draft me an NDA" to a review-ready, Word-exportable document in minutes, not just a block of generated text. Drafting became a continuous workflow: chat → document → review → redlining → follow-ups AI stopped being a starting point and became part of the full drafting lifecycle.
• AI usage increased because it aligned with existing legal workflows • Review and redlining happened directly in the document • Final outputs matched what lawyers were confident sending externally
The document-first workflow became a key reason customers chose Wordsmith. We won deals where we were more expensive — because competitors couldn't do this.
This is where Wordsmith stopped being 'AI for lawyers' and became a real drafting tool.
Wordsmith
Designing a Slack-native AI system for managers
From everyday work signals to calm, actionable performance insights
My focus was designing how AI shows up inside Slack—turning messy, real-world team activity into insights managers could trust, act on, and come back to.
Problem
Managers need performance insights they can trust and act on — but team activity is messy and scattered across conversations and tools.
Slack is where work actually happens, yet most AI and analytics live outside it. As a result, insights feel disconnected from the flow of work.
Managers needed signals turned into calm, actionable insights inside Slack — without another dashboard or context switch.
Work was happening in Slack — but insight and support were not.
Challenge
How might we design AI that shows up naturally inside Slack — turning real team activity into insights managers could trust, act on, and return to?
This wasn't just about prompts. The challenge was bridging two worlds:
- Web UX thinking — structure, hierarchy, clarity
- Conversational Slack experiences — tone, timing, feedback loops
Designing AI behaviour — not just messages.
Outcome
We designed how AI shows up inside Slack — from prompts and response patterns to message structure and feedback loops managers could rely on.
The system turns everyday work signals into calm, actionable performance insights, designed for the context where managers already are.
- AI prompts and response patterns for managers in Slack
- Clear message structure, tone, and feedback loops
- Web UX principles translated into conversational experiences
AI became a quiet partner in day-to-day management — not another dashboard to maintain.
Performance support became part of real work — not a separate process.
LLM-driven Slack messaging system
Designed message formats, prompts, and conversational structure to ensure AI insights feel calm, relevant, and human inside Slack.
Shaped how Oli introduces itself, asks questions, and supports managers by defining tone, intent, and guardrails for AI-generated messages.
Manager home & Slack entry point
A control surface that helps managers review insights, invite teammates, and move seamlessly into Oli's Slack conversations without breaking flow.
Team connection & context setup
A lightweight setup flow to connect managers, direct reports, and peers — giving Oli the relational context it needs to personalise insights and coaching.
Team relationships management
A personalized home where managers review insights, invite teammates, and continue with Oli in Slack—with a clear view of their team structure and collaboration.
Integrations & data sources
Connect tools like Slack and Google Drive so Oli can bring clearer insights from meetings and work—with clear status and available integrations.
Slack-native AI conversations
Designed the structure, formatting, and interaction patterns of Oli's Slack messages—shaping how AI insights appear in real team conversations.
Onboarding & coaching tone design
Defined message tone, pacing, and prompts in collaboration with LLMs to deliver supportive, trustworthy guidance without sounding prescriptive or intrusive.