Víctor Jean Pierre Mandujano Gutierrez. Senior software engineer and Tech Lead with ten years building and leading production systems, now working agent first: I run the whole development lifecycle (specification, implementation, review, verification) with orchestrated AI agents, and I help engineering teams adopt them within whatever policy their organization allows.
Geological engineer from UNI, which is why half my work sits in computational geoscience and mining.
Deep Skill is the company I founded and run as an AI-native organization: a small senior team plus agents doing the work of a much larger one. It has two lines of business.
|
A learning platform for science and engineering, with an AI coach. Structured paths in algorithms and data structures, mathematics, data science (Python, SQL, data engineering, machine learning), AI-assisted software development and system design. The AI coach follows each learner through practice, feedback and progress, on top of instructors who are ICPC world finalists and Google engineers. For companies: Corporate AI Programs, intensive tracks that turn developers, engineers and analysts into people who build with AI. |
Automation and AI-supported operations for companies. Turbo, our AI software factory: agents orchestrated by senior engineers, shipping in continuous increments instead of one-off projects. Typical work: workflow and back-office automation, data pipelines and system integrations, legacy modernization and migrations, and LLM features in production with the evaluation, observability and guardrails they need to stay there. |
Also Fractional Tech Lead at Rumbo. Before that: API Tech Lead and Engineering Manager at Scotiabank's Digital Factory, and founder and CTO of Deep Pit Technology, acquired by Stracon Tech.
Every contribution below was specified, generated and reviewed by me before it went upstream. The agents write; I own the design and the verification.
Merged upstream
| Infrastructure and developer tooling | Computational geoscience |
|---|---|
|
|
|
Not merged, but the fix that landed builds on them
prometheus/client_java · #2316 → #2396
Summary quantiles collapsing to the minimum: the merged fix shares this PR's two-part diagnosis, the compress() merge bound and the early-stopping scan in get()
DASDAE/dascore · #927 → #1290
Times outside the nanosecond range: the maintainer's smaller fix reuses this PR's test cases and credits it
In review
apache/lucene #16643 · keycloak/keycloak #52478 · OpenAPITools/openapi-generator #25073 #25074 · wailsapp/wails #6205 #6206 · moghtech/komodo #1661 #1662 · Automattic/harper #4425 · dapr/python-sdk #1210 · testcontainers-java #12072 · torchgeo #4111 · gempy #1110 · fatiando/harmonica #698
I also review upstream, not just submit: testcontainers-go and dascore among others.
open-source-builder is the agentic system behind the list above. Claude Code agents scan GitHub for active projects in my domains, filter issues by maintainer responsiveness, AI policy and real reproducibility, and hand me a curated report. The contributions are the output; this is the method.
