Ph.D. candidate · FEUP × CMU Portugal

Making formal proofs usable.

I'm Álvaro, a software-verification researcher in Porto. I build tools that make verification-aware languages like Dafny easier to trust and to work with. My current focus is checking whether a verified spec is actually strong enough: ProofPulse (ASE 2026) shows which parts of a specification a proof really depends on, exposing vacuous proofs and unnecessary preconditions. I also work on LLMs that suggest the missing assertion and debuggers that point to the line where a proof broke. I'm advised by Alexandra Mendes (FEUP / INESC TEC) and Ruben Martins (Carnegie Mellon).

Portrait of Álvaro Silva wearing glasses and a light-blue shirt
fig. 1 — porto, pt
research.dfy
method Research(dev: Developer) returns (p: Proof)
  requires dev.writes("Dafny")
  ensures  p.verified && p.understood
  ensures  dev.effort < old(dev.effort)
verifying… 1 verified, 0 errors
ProgramCMU Portugal Affiliated Ph.D.
FocusSpec strength · ProofPulse · Dafny
Before3 yrs ASIC verification at Synopsys
AlsoGuest assistant at FEUP
01 / Changelog

Lately

  1. ASE 2026DSpec2Test: tests generated from Dafny specifications alone, which kill 93.9% of mutants.
  2. ASE 2026ProofPulse: proof-coverage analysis that flags specifications which verify but say little.
  3. ISSRE 2026Automated fault localization for Dafny will appear at the 37th IEEE ISSRE.
  4. serviceReviewer for FM 2026 and artifact evaluator for FormaliSE 2026.
  5. SEFM 2025Can LLMs help students prove software correctness? published in the SEFM 2025 proceedings.
  6. preprintDAISY: LLMs that infer several missing helper assertions in Dafny at once.
02 / Research

Four threads, one question

What would make a working developer reach for a verifier, and keep using it?

ensures true i

Specification quality

A proof that verifies can still rest on a weak spec. My main focus is ProofPulse (ASE 2026), which measures which spec elements a proof actually relies on and flags vacuous proofs and unnecessary preconditions. DSpec2Test (ASE 2026) complements it by turning the spec itself into a test suite.

ASE '26ProofPulseDSpec2TestMutDafny
assert ? ii

LLM-assisted proof

Language models that fill in the tedious parts of a proof, such as helper assertions and lemmas, while the verifier checks every suggestion. I also study how students use these assistants when they write proofs.

DAISYFormaliSE '24SEFM '25
✗ → ✓ iii

Debugging & repair

When verification fails, I want the tool to say where and why. This thread covers fault localization from counterexamples, single-transformation repair, and the FCT-funded VeriFixer project with INESC TEC, INESC-ID and CMU.

ISSRE '26APR @ ASE '26DafnyFixVeriFixer
rtl ∪ rust iv

Beyond Dafny

I spent three years verifying hardware with SystemVerilog, UVM and VCFormal, and I want to bring that experience back into the work. Current interests are Dafny as a language for hardware models, Verus for verified Rust, and agent-aware editors such as TraceLean.

VerusSystemVerilogTraceLean IDE
03 / Selected papers

Publications

Full list with abstracts and links →

  1. ASE2026

    ProofPulse: Interactive Proof Coverage Analysis for Dafny

    Álvaro F. Silva, Ruben Martins, Alexandra Mendes

  2. ISSRE2026

    Improving Debugging in Verification-Aware Languages Through Automated Fault Localization: A Case Study in Dafny

    Álvaro Silva, Isabel Amaral, João Pascoal Faria, Alexandra Mendes

  3. ASE2026

    DSpec2Test: Specification-Driven Test Generation in Dafny

    Sofia Vieira Pinto, Álvaro F. Silva, João Pascoal Faria, Alexandra Mendes

  4. arXiv2025

    Inferring Multiple Helper Dafny Assertions with LLMs

    Álvaro Silva, Alexandra Mendes, Ruben Martins

  5. SEFM2025

    Can Large Language Models Help Students Prove Software Correctness? An Experimental Study with Dafny

    Carolina Carreira, Álvaro Silva, Alexandre Abreu, Alexandra Mendes

  6. FormaliSE2024

    Leveraging Large Language Models to Boost Dafny's Developers Productivity

    Álvaro Silva, Alexandra Mendes, João F. Ferreira

04 / Path

From silicon to proofs

Experience

  1. Sep 2024 — now

    Ph.D. Candidate

    FEUP, affiliated with Carnegie Mellon University
    • CMU Portugal Affiliated Ph.D. Program grant.
    • Usability of verification-aware languages, with a focus on Dafny.
    • Researcher on the VeriFixer project.
  2. Sep 2024 — now

    Guest Assistant

    FEUP
    • Functional & Logic Programming, Programming Fundamentals (Python), Computing Systems.
  3. Aug 2021 — Sep 2024

    ASIC Digital Design Staff Engineer

    Synopsys · Porto
    • Verification lead for MIPI CSI-2 camera controllers.
    • Functional and formal verification (VCFormal) of automotive IP; ISO 26262 certified.
  4. Oct 2020 — Aug 2021

    Researcher

    INESC TEC · Porto
    • European project TRUST-AI; two research scholarships.

Education & awards

  1. 2016 — 2021

    Integrated M.Sc. in Electrical & Computer Engineering

    FEUP · final grade 19/20
    • Specialization in Microelectronics, minor in Telecommunications.
    • Thesis: interpretable dispatching rules for job-shop scheduling with dimensionally aware genetic programming.
  2. 2019 — 2020

    Erasmus exchange semester

    Technische Universität München · GPA 1.27
  3. 2017 — 2018

    Merit scholarships & Incentivo award

    FEUP
    • Merit scholarships in 2016–17 and 2017–18.
    • Incentivo 2018, awarded to the top 20 first-year students.
05 / Community

Teaching & service

Teaching

FEUP guest assistant

  • Functional and Logic Programming
  • Fundamentals of Programming (Python)
  • Computing Systems
Service

Program & artifact committees

  • Reviewer, Symposium on Formal Methods (FM 2026)
  • Artifact Evaluator, FormaliSE 2026
Mentoring

Thesis supervision

  • Industry supervisor of R. P. de Jesus, Evaluating High-Level SystemC Formal Verification Compared to RTL Verification (2024).
06 / Contact

Let's prove something together.

I'm happy to talk about Dafny, Verus, program repair, LLMs for verification, or hardware verification, whether you have a collaboration in mind or just a question.