Jose Aldave
Senior Product Designer · AI products

I design AI products that feel simple, and I build them myself.

Five years at Torre.ai, a US-based AI job network, where I grew from recruiting operations to designing and shipping its AI agent and consumer experience. Based in Peru, working remotely with US teams.

Conversational AIMarketplaces & matchingCode-backed prototypesClaude Code · Cursor · Figma MCP
See Emma, live ↓
Jose Aldave at his desk, with Torre.ai open on his laptop
25% → 95%Emma's success rate, in one month after launch
8× fasterFull platform redesign, against the original estimate
2 minReply time for 1,200–1,500 candidate conversations a day

Selected work

Three case studies from Torre.ai, where I owned the problem from framing to production.

Emma's live profile on Torre.ai: Torre's AI talent agent, with her skills, reputation and a Message button
Emma's live profile on Torre.ai
Try it yourselfTell Emma what job you're looking for and see how she searches for you.
Chat with Emma ↗
Case 01 · Conversational AI

Emma: one friendly face, eight agents behind it

Torre.ai2022 – 2026Role: product & agent design

The problem

Torre matched candidates and jobs with an algorithm, but reaching candidates, answering their questions and interviewing them still depended on people. A team of about 15 answered 1,200 to 1,500 messages a day with a two-minute response target.

What I did

  • Co-designed the agent architecture with an engineer: about eight specialized agents that pass work to each other, behind a single conversation the user sees.
  • Wrote the prompts and function calls, and the knowledge base Emma consults, including when she should consult it.
  • Placed rule-based checks around the parts where the model was inconsistent, and kept people in the loop until quality held.

Decisions and trade-offs

  • One conversation, many agents. A single large prompt answered the same message differently each time. Splitting the work let us test and fix each step on its own.
  • People in the loop first. Full automation on day one would have put candidates at risk. For about two months, reviewers checked answers in shifts while we raised quality.

Impact

25%success rate at launch
75%three days later
95%within a month, replacing manual replies with real-time answers

Emma later grew to run candidate interviews, ask for references, help companies post jobs and learn each user's preferences.

Case 02 · Visual redesign

From dark to light across eight services, 8× faster than estimated

Torre.ai2026Role: design lead & implementation

The problem

Moving the whole platform from a dark to a light theme had waited in the backlog for years. It touched about eight services, each with its own interface, and was estimated at more than six months.

What I did

  • Mapped the repositories of every service and defined the new palette, checking contrast on every surface.
  • Gave the interface a soft, tactile look and documented components in Storybook so the change stayed consistent.
  • Implemented it myself with Claude Code and Figma, tested every flow locally, and sent pull requests that engineers reviewed before release.

Decisions and trade-offs

  • Build it, don't hand it off. Static mockups for eight services would have meant months of back and forth. Working in code with AI let me design and ship in the same loop.
  • System before screens. Defining the palette and components first kept eight different interfaces looking like one product.

Impact

8× fasterthan the original estimate of more than six months
Case 03 · Untangling a complex product

One place to find everything Torre can do

Torre.ai2026Role: product & design, end to end

The problem

Torre had many useful features for companies and recruiters, but they were scattered. People didn't know they existed or how to reach them.

What I did

  • Identified the problem and made the case for it, then designed a single hub that brings every feature into one guided flow.
  • Connected the hub to the existing services and wrote the architecture decision records for the changes it needed.
  • Built and shipped it, from design to production.

Decisions and trade-offs

  • Guide, don't list. A menu of every feature would have repeated the problem. The hub asks what the user wants to do and takes them there.

Impact

Raised satisfaction among companies and recruiters using the platform, and made features they already paid for discoverable.

Side project · Consumer web app

Nivio, a tennis web app I designed and shipped on my own

Personal2026Live web app

I play tennis, so I built a web app for players to log matches and sets and see the ranking in their city. Tournaments are next. It started as a way to validate the idea before moving to the app stores.

My loop: Claude builds a working preview, I review it as a user and as a designer, and only then implement. It is the same workflow I use for client work.

9:41
Nivio sign-in screen: email or nickname field and Continue with Google
9:41
Nivio home screen: log a game or a training session, log with AI, pending items and recent activity
Sign-in and home screens, live at getnivio.app
Try it yourselfOpen it in your browser, no download needed. It works best on your phone.
Open Nivio ↗

How I work

From a vague problem to a shipped flow, without stopping at static mockups.

01Frame the problem

Talk to the people who live it, map the current journey, and find where it breaks.

02Prototype in code

Use Claude Code or Cursor to get a working version in hours, so we test behavior, not pictures.

03Refine the craft

Bring it into Figma through MCP, apply the design system, and polish every state and detail.

04Ship and measure

Pull requests reviewed by engineers, then track the metric the work was meant to move.

ClaudeClaude CodeCursorFigmaFigma MCPStorybookGit & GitHubBPMN

How I got here

I started on the operations side, which is why I design from how people actually work.

RecruiterTorre.ai, 2021
Trainer & process designOnboarded recruiters, mapped processes in BPMN
Service designDefined how clients are served
Product designProductized those services into self-service
AI Product ManagerReporting to the CEO, until 2026

Business Administration, Universidad Nacional de Trujillo, graduated second in a class of about 100. Fluent in English and Spanish.