╭──────────────────────────────────────────────────────────────╮
│   o'')}____//   lluc-code v1.0                               │
│    `_/      )   Lluc's agentic profile · hacker-mode: on     │
│    (_(_/-(_/    cwd: ~/lluc/projects                         │
╰──────────────────────────────────────────────────────────────╯

> /whoami --deep

● I'll read Lluc's repos and keep only what actually matters.

● Read(prosper-ai/role.md)
  ⎿  AI Engineer @ Prosper AI (YC, $30M Series A, a16z).
      Voice AI agents for US healthcare, thousands of
      calls/day. Reliability, integrations with 80+ EHRs.

● Read(divergence/README.md)
  ⎿  Divergence check for Scala 3 match types, implemented
      in the compiler (scala/scala3 #24661). Endorsed by
      Martin Odersky, now in the SIP process (PR #136).

● Read(lausanne/equity.md)
  ⎿  Pitched Melexis and Magyar Telekom into a 25k CHF
      real-money fund, 75%+ of the votes. Up 28% and 53% in
      8 months.

● Grep("hackathon", ~/lluc/projects)
  ⎿  YC Voice Agents Hackathon: ClearPath, a Pipecat voice
      agent for medical intake. SUI Hackathon 2025: $1,000
      prize winner (Healix).

● Done reading. Here's what actually matters:

AI Engineer at Prosper AI (YC, $30M Series A led by a16z)

I build voice agents that talk to real people, in real time, and don't fall over. At Prosper that means thousands of healthcare calls a day: sub-second latency budgets, barge-in that actually works, telephony that misbehaves, and an LLM that has to be right the first time because there is a patient on the line.

Before this I graduated first of ~400 in Informatics Engineering at UPC, spent a semester at UC Irvine as a Balsells Fellow doing cooperative game theory under Vijay Vazirani, and an exchange semester at EPFL. I like problems where the theory is hard and the demo still has to run.

Voice & real-time

Backend

Interfaces

Most of what I know, I learned by shipping something that had a deadline.

• ClearPath · YC Voice Agents Hackathon. A Pipecat agent that replaces medical intake paperwork by filling a live React form as the patient speaks. Hardened with simulated-caller evals.
• divergence · Scala Improvement Process endorsed by Martin Odersky (creator of Scala, author of javac) to define the behavior of match types (compiler PR). Catches 100% of infinite reductions at compile time, backed by a termination proof generalizing Dickson's Lemma, and keeps the full Scala 3 test suite compiling unchanged.
• leximin · research to code. First public implementation of Vazirani's ACM EC algorithm for leximin-fair profit allocation. Exact rational arithmetic, cross-validated on 1,320 instances.
• Healix · SUI Foundation Hackathon $1,000 prize winner. An encrypted mood journal on the Sui blockchain.
• PyCOMPSs LLM · Barcelona Supercomputing Center. Fine-tuned CodeLlama 7B with LoRA on MareNostrum 5 into the first LLM that writes PyCOMPSs parallel code, plus the tested pipeline that built its training datasets from 30+ real HPC applications.

Daily calisthenics, plus running and cycling outdoors. I read science fiction, finance and self-improvement, and Stephen King outranks them all. The rest goes to side projects with whatever the state of the art shipped this month.

Humans ran on voice for a hundred thousand years; screens got the last fifty. Every form, menu and dashboard is a workaround for the same limitation: software couldn't hold a conversation. That limitation just died, and most of the industry is still building workarounds.

I want to make the interfaces that shouldn't exist disappear, which is almost all of them. You say what you need, the machine does it, and it feels less like operating a tool and more like talking to a competent colleague. I'm starting where it's hardest and matters most: real-time, real phone lines, real patients.

If you're building toward the same future, email me.

_{Barcelona, Spain}