Hugo Vicente
Technical Product Manager · Software Engineer · Solutions Engineer · Forward Deployed Engineer
CS Engineering graduate from Instituto Superior Técnico. I build and ship distributed systems, ML pipelines, and customer-facing platforms end to end — then turn that same engineering judgment into product decisions, on weekly release cycles.
About
I'm a Computer Science & Engineering graduate from Instituto Superior Técnico, currently a Product Manager Intern on Cloudflare's Tunnel team.
I sit at the intersection of product and engineering: translating ambiguous stakeholder requirements into modular systems, standing up the infrastructure to run them, and shipping customer-facing platforms on weekly release cycles. Recently I led the technical roadmap for a synthetic clinical-data platform, cut model experimentation costs by up to 70%, and delivered an AR education platform for the Calouste Gulbenkian Museum. I also own the architecture of TraceBloom, an LLM observability platform spanning a Rust ingestion service, ClickHouse trace storage, and a live-streaming dashboard — including its security hardening.
Whether the work is product strategy, an ML pipeline, or 80+ Linux servers, I care about the same thing: turning complexity into something people can actually use. I'm looking for roles where product and engineering overlap — and I do my best work in teams that ship.
Experience
-
Product Manager Intern Current
Cloudflare · Tunnel
- Partner with engineering, design, and customer-facing teams to scope and prioritize roadmap items for Cloudflare Tunnel, translating technical constraints into product requirements.
- Contribute to cross-functional execution on Tunnel product initiatives, working within Cloudflare's technical product development process.
-
Project Manager
JUNITEC
- Owned the technical roadmap for a synthetic clinical-data platform, translating ambiguous stakeholder requirements into a modular ML pipeline combining CTGAN, TimeGAN, Synthea, and BioGears.
- Cut model experimentation costs by up to 70% by standardizing GPU training on Google Colab Pro and designing a "train remotely, infer locally" workflow.
- Managed pipelines over high-frequency biosignal datasets (500 Hz) from 6,000+ surgical patient records, coordinating preprocessing, training, and validation.
- Established weekly delivery checkpoints across engineering and research workstreams, improving execution visibility.
-
Software Developer
JUNITEC
- Developed an interactive AR-based educational platform for the Calouste Gulbenkian Museum using React and Firebase.
- Translated stakeholder requirements into weekly engineering deliverables across a multidisciplinary team of designers and developers.
-
System Administrator
RNL · Instituto Superior Técnico
- Administer and monitor 80+ Linux servers, including HPC clusters and virtualized environments, supporting students, researchers, and faculty.
- Built Grafana observability dashboards, reducing incident diagnosis time by 35%.
- Automated infrastructure monitoring and operational workflows, improving system visibility and reducing manual debugging overhead.
Projects
TraceBloom
Directed AI agents (architecture, implementation, review) to build an LLM observability platform: TypeScript/Python SDKs, a Rust OTLP collector built for backpressure so ingestion never falls behind, ClickHouse for write-heavy trace storage at scale, a Next.js dashboard streaming live traces over SSE, and an idempotent eval engine that flags prompt regressions without re-scoring. Owned every architecture tradeoff across 7 milestones (31 decisions documented), enforced 170+ automated tests in CI, hardened the public demo with defense-in-depth (network isolation, read-only DB credentials, zero real API keys), and remediated 14 dependency CVEs plus an IP-spoofing gap before going public.
KVS Subscription Server
Multithreaded key-value store in C with a named-pipe client-server architecture and real-time subscription notifications. POSIX synchronization via mutexes, semaphores, and read-write locks to preserve consistency across concurrent sessions.
Fault-Tolerant Blockchain
Distributed replicated blockchain ledger with a central sequencer for global transaction ordering and RSA-2048 signed blocks across nodes. Optimistic execution, linearizable reads, and client failover for low-latency, fault-tolerant operation.
More on github.com/hugo-vicente11 ↗
Skills & Toolkit
Programming Languages
- Python
- C
- Java
- C++
- SQL
- Rust
- TypeScript
Tools & Tech
- Git
- Linux
- Docker
- React
- Firebase
- PostgreSQL
- Streamlit
- Pandas
- Plotly
- Grafana
- Google Colab
- Wireshark
Certifications
- Data Science with Python
- Machine Learning
- Intro to Quantum Computing — IBM
Languages
- Portuguese — native
- English — fluent
- Spanish — basic
- French — basic
Education
B.Sc. in Computer Science & Engineering
Instituto Superior Técnico — Lisbon, Portugal
Get in touch
I'm always open to conversations about product, engineering, and building things that ship. The fastest way to reach me is email.