01 / 07 — Context

Digital transformation and Peru's strategic opportunity

In a world driven by accelerated digital transformation, Artificial Intelligence has become a key force of innovation and national development. Peru stands at a critical juncture: it must transition from being a passive consumer of foreign technology to an active creator of knowledge, solving local problems with global standards and fostering a new era of technological sovereignty.

Peru must transition from a technology consumer to a knowledge creator through AI adoption.
According to the Economic Commission for Latin America and the Caribbean (CEPAL, 2025, p. 142), Peru leads the region in web-based AI usage intensity (6.97 visits per user) and ranks third in generative AI adoption (74.6 points), demonstrating high user engagement. This adoption intensity indicates a willingness to leverage AI for productivity gains and localized problem-solving, evidenced by Peru's leadership in both general AI platform web traffic (100 points) and advanced usage (1.19 visits per user). However, this transition is probable but conditional on sustained investment in infrastructure and talent development. Without addressing the zero participation in top-tier conferences (NeurIPS, ACL), Peru risks remaining a consumer of foreign models despite its high adoption rates (Soto et al., 2025, p. 95).
1st
in web AI usage in the region
3rd
in generative AI adoption
0%
participation in top conferences
References Soto, Á., Durán, R., Moreno, A., Adasme, S., Rovira, S., Jordán, V., & Poveda, L. (Coords.). (2025). Índice Latinoamericano de Inteligencia Artificial (ILIA) 2025.
02 / 07 — Community

Collective research and continuous learning

We explore AI systems through rigorous experimentation, reverse engineering, and continuous collaborative learning. Our philosophy is simple: the goal is not to copy, but to understand. By dissecting how models work, we adapt and evolve them to suit our unique cultural and economic context, ensuring innovation is a shared community effort.

Collective research is essential to develop contextually relevant AI solutions in Peru.
Only 75 researchers consistently publish on AI in Peru (2.23/million, 10th place), with 0% participation in top-tier conferences (Soto et al., 2025, p. 92). Fragmented research limits Peru's capacity to develop local models and solutions: Brazil and Mexico concentrate 68% of active AI researchers in Latin America (Soto et al., 2025, p. 28). Collective research can succeed with coordinated academic-industry partnerships. Language barriers and limited access to computational resources can hinder collaboration (Soto et al., 2025, p. 100).
75
active AI researchers
10th
regional rank (2.23/million)
68%
concentrated in Brazil and Mexico
References Soto, Á., et al. (2025). Índice Latinoamericano de Inteligencia Artificial (ILIA) 2025.
03 / 07 — Infrastructure

Open-source infrastructure

Open technologies are the foundation of transparency, accessibility, and true innovation capacity. We leverage Kubernetes and open ecosystems to deploy scalable AI solutions. By integrating LocalAI for API compatibility and AnythingLLM for document management, we create robust, self-hosted environments, ensuring researchers in remote areas of Peru have the same tools as major global tech centers, without relying on external cloud providers.

KubernetesLocalAIAnythingLLMSelf-hosted
Open source is the most viable path for Peru to develop AI sovereignty.
Honduras, El Salvador, and Cuba stand out in open-source productivity, quality, and relevance (Soto et al., 2025, p. 28). Open source reduces dependence on proprietary systems and enables local customization. High performance in open-source production indicates strong collaborative development communities (Soto et al., 2025, p. 113). Success depends on sustained investment in training and infrastructure. Open source requires technical expertise that may be scarce in remote areas (Soto et al., 2025, p. 100).
References Di Giacinto, E. (n.d.). LocalAI. Mintplex Labs. (n.d.). AnythingLLM. Soto, Á., et al. (2025). ILIA 2025.
04 / 07 — Practice

From theory to application

We go beyond theoretical discussion, implementing state-of-the-art transformer architectures and generative models using PyTorch, Diffusers, and ik_llama.cpp. Our workflow includes llama-factory for efficient fine-tuning of models on local datasets, and wllama.cpp (WASM) to run these models directly in the browser. This practical approach enables deep understanding of internal mechanisms without relying on external APIs, giving developers the tools to create robust solutions for education, health, and agriculture within Peru.

PyTorchDiffusersik_llama.cppLLaMA-Factorywllama (WASM)
Practical implementation of AI tools is critical to closing the gap between theory and real-world impact.
Peru leads AI web traffic with 100 points (6.97 visits/user) and 1.19 advanced AI visits (1st place) (CEPAL, 2025, p. 142). High web traffic indicates strong user engagement with AI tools, creating opportunities for application development. 15 out of 19 ILIA countries show good results in app development, indicating it is an accessible entry point for AI adoption (Soto et al., 2025, p. 28). Implementation success depends on infrastructure availability and developer skills. Limited GPU access (only 7,235 TF/s HPC, 3rd national place) may restrict model training (CEPAL, 2025, p. 48).
References Kawrakow, I. (n.d.). ik_llama.cpp. Zheng, Y. (n.d.). LLaMAFactory. Soto, Á., et al. (2025). ILIA 2025.
05 / 07 — Identity

AI sovereignty for Peru

Our ultimate goal is to build AI systems that speak our language — literally and culturally. We are training models specifically for Peruvian Spanish and local contexts, including multilingual support vital for indigenous languages like Quechua and Aymara. This ensures AI serves all Peruvians, preserving and elevating our linguistic heritage.

AI sovereignty through local linguistic models is essential for cultural relevance and inclusion.
LatamGPT prioritizes Spanish, Portuguese, English, and indigenous languages like Quechua (CEPAL, 2025, p. 58). Local models prevent cultural biases and ensure relevance for regional needs. 80% of top global models are trained solely on English data, creating a gap for non-English contexts (CEPAL, 2025, p. 57). Sovereignty is achievable with regional collaboration and sustained investment. Data scarcity and computational requirements may limit model quality (CEPAL, 2025, p. 100).
80%
of top models trained only in English
4+
priority languages in LatamGPT
References Soto, Á., et al. (2025). ILIA 2025.
06 / 07 — Invitation

Join the movement

AI made in Peru, for Peru, requires a community of passionate minds. We formally invite researchers, students, universities, and institutions to collaborate with us. Together, let us build a sustainable, ethical, and sovereign AI ecosystem that drives Peru toward the future.

Collaborative AI development is essential for Peru's technological sovereignty.
Foundations: Peru's 1.12% of global AI investment and only 6 unicorn companies indicate significant underinvestment (Soto et al., 2025, p. 29). Collaboration can pool resources and expertise to overcome investment gaps. Regional initiatives like LatamGPT demonstrate successful multi-institutional collaboration (CEPAL, 2025, p. 57). Success requires institutional support and policy alignment. Fragmentation and lack of coordination can hinder collaborative efforts (CEPAL, 2025, p. 153).
Visit official platform →
References Soto, Á., et al. (2025). Índice Latinoamericano de Inteligencia Artificial (ILIA) 2025.
07 / 07 — Roadmap

Prototypes bridging phases: Infrastructure → Applications → Training

This project is sponsored by Colegio Santa Rosa de Lima, San Borja and operated by its Artificial Intelligence Ethics Committee, an autonomous technical body guiding the ethical, safe, and sovereign adoption of AI. Our roadmap does not depend on the state calendar, but responds to internal technical and pedagogical logic: each prototype simultaneously crosses all three phases (infrastructure, applications, training), improving incrementally and progressively. For example, the pedagogical management content creation assistants prototype requires: (1) local infrastructure (Kubernetes, LocalAI, AnythingLLM, ik_llama.cpp) to run models in the classroom; (2) concrete application (ethical validation exercise generation interface); and (3) teacher training (workshops on prompts and human oversight). Each improvement in one phase drives improvements in the other two —without waiting for external decisions or relying on external infrastructure.

The iterative and incremental approach, with prototypes bridging infrastructure, applications, and training, is viable and ethically sustainable, aligned with the National AI Strategy but autonomous in execution.
The National AI Strategy 2026–2030 (ENIA) establishes general principles for the ethical use of AI, including transparency, human oversight, and data protection (PCM, 2025a, p. 45). Law No. 31814 explicitly recognizes that "educational institutions may develop their own AI initiatives, provided they align with national ethical principles" (Art. 17, inc. 4), granting technical autonomy to design independent routes (Peru, 2023, p. 8). Our route prioritizes local infrastructure (Kubernetes, LocalAI, AnythingLLM, ik_llama.cpp) before applications, because without sovereign infrastructure there is no technological sovereignty (DGCC, 2025a, p. 100). The Santa Rosa de Lima College AI Ethics Committee will validate each phase through ethical impact and privacy assessments, in accordance with NTP-ISO/IEC 23894:2023 (INACAL, 2023, p. 7), without relying on external approvals. This iterative and incremental approach avoids state fragmentation: while the government defines frameworks, we execute validated pilots before scaling (ENIA, 2025b, p. 59). The Assessment of Peru's AI Readiness Stage (UNESCO, 2025, p. 31) confirms the country needs to address infrastructure, talent, and funding simultaneously — our methodology does so from the start. The Santa Rosa de Lima College Internal Regulations 2026 establishes in Art. 152° the creation of the AI Ethics and Data Protection Committee, with technical and operational autonomy to issue binding opinions (Santa Rosa de Lima College, 2026, p. 152).
Phase 1
Physical infrastructure (Kubernetes + LocalAI + AnythingLLM + Ik_llama.cpp)
Phase 2
Applications: Pedagogical management content creation assistants, virtual tutors
Phase 3
Training: Teacher workshops, ethics workshops, validated prototypes
ENIA 2026–2030Law 31814ISO/IEC 23894:2023INACALDGCCInternal Regulations 2026
References Santa Rosa de Lima College. (2026). Internal Regulations 2026. INACAL. (2023). NTP-ISO/IEC 23894:2023. PCM. (2025a). ENIA. PCM. (2025b). Regulation Law 31814. Peru. (2023). Law 31814. DGCC. (2025). AI Strategy. UNESCO. (2025). Readiness Assessment.
1 / 7
Loading WASM engine and Embeddings...
La Golondrina
🤔
Thinking...

La Golondrina Cuántica

Sovereign AI · Semantic RAG (Transformers) · 20 Cards

Voice Terminal
Waiting for model...
Initializing system. Loading local libraries...