Latin American hyperlocal social marketplace and community platform came to Armakuni for openAI-to-AWS Bedrock migration, voice-first AI interfaces for low-literacy users, conversational onboarding, AI content moderation, and AI commerce companion.
The challenge
Armakuni won the engagement by demonstrating a clear roadmap that addressed A Latin American hyperlocal social marketplace and community platform' cost structure and prepared the platform for 1-2 million users. The assessment SOW validated the architecture before implementation began, and the migration SOW executed the three-phase build. Armakuni now owns the AI architecture that runs the entire platform.
Scale: 143,000+ active users across Argentina and Costa Rica; targeting 1-2 million users across Latin American markets.
What we built
Armakuni delivered openAI-to-AWS Bedrock migration, voice-first AI interfaces for low-literacy users, conversational onboarding, AI content moderation, and AI commerce companion. The engagement ran as two fixed-scope SOWs: Assessment Phase SOW (August 2025) and AWS SaaS Migration & Enhancement SOW (November 2025). The work was AWS funded.
The outcomes
6 measurable outcomes shipped across the engagement. The ones that moved the business the most:
Every AI workflow migrated from OpenAI and LLaMA to AWS Bedrock. Conversational, moderation, and onboarding workflows now run on AWS, and AI cost is under the platform's control, optimized by model selection rather than a fixed per-call rate.
A voice-first interface is live and validated with lower-literacy users. The platform now reaches an audience that form-based interfaces excluded entirely, opening markets where voice is the primary interaction mode rather than an accessibility feature.
Conversational onboarding now replaces forms. Users complete registration through natural-language dialogue rather than data entry, addressing drop-off at registration at the architectural level rather than through UI iteration.
AI content moderation with positive rephrasing now runs at scale. Community quality holds without human review as volume grows, and a feedback loop improves accuracy with every interaction rather than requiring manual rule updates.
Architecture validated for 1 to 2 million users before the Mexico City launch. Auto-scaling policies are in place and AWS cost is modeled against real usage, so the business enters its biggest launch with infrastructure confidence rather than uncertainty.
An AI companion is live with memory and preference learning. It builds user models that improve every session, positioning the platform to activate autonomous commerce features without a foundational rebuild when the product is ready.
Built on AWS
The production environment runs on 9 first-party AWS services. Delivered under our AWS AI Services and SaaS Consulting competencies.
What's next
Autonomous commerce activation , enabling the AI companion to facilitate purchases and transactions directly through conversational interfaces, building on the foundation laid in Phase 3.