I’m Ramiro Carnicer Souble. Ask the assistant anything about my experience, projects or how I work.
AI-native product for People Analytics and organizational intelligence. Goes beyond dashboards — it predicts, signals, and acts.
Hackathon project that placed 2nd on the leaderboard and was selected for an accelerator. Built and shipped as a focused product sprint.
Competitive platform where autonomous agents play poker via REST API. Built around game loop architecture, agent isolation, and real-time state management.
Community-powered LLM inference. Access specialized, community-hosted models through an OpenAI-compatible API — cheaper and more flexible than cloud defaults.
Portable, spec-driven development kit with multi-agent orchestration. Turns technical specs into implemented code using a structured pipeline of specialized agents. Presented as a talk on SDD + multi-agent architecture at the first Codex Argentina meetup.
Scalable APIs, microservices, and distributed architecture. Reliability is not optional when the systems need to hold under real load.
Agent workflows, applied LLMs, and productized AI. The interesting part is making these systems useful enough to deploy.
From zero to shipped MVP. Technical judgment and product taste are part of the same craft.
Architecture decisions, stack selection, and pragmatic tradeoffs for products that need to move fast without collapsing later.
Backend Software Engineer at Mercado Libre
Scalable backend systems at Latin America's largest e-commerce and fintech platform.
A practical take on choosing the least complex orchestration pattern that solves the task — from augmented LLMs and routing to orchestrator-workers and autonomous agents.
Reposted from @ram4_dev: Compass was a hackathon project, finished 2nd on the leaderboard, and was selected for an accelerator.
On building MVPs, confronting market silence, and iterating without ego. The loop that actually works.
Data reporting is not strategy. Real analytical value is forward-looking — signal detection, not passive charts.
The real bottleneck isn't tooling — it's the mental model. A systems-thinking take on HR and organizational intelligence.
Software is craft. I care about systems that work in the real world — not just in demos. Technical depth is leverage, not identity theater.
Shipping beats planning theater. I prefer a live product with rough edges over a perfect spec with no users.
I want to understand the product vision and the technical constraints at the same time. Engineers who only see code miss half the problem.
I’m interested in AI when it connects to real workflows and changes how work actually gets done.
Choosing the right abstraction and choosing the right feature are closer than most people think.
If you're building something ambitious in AI, backend infrastructure, or early-stage product, I’m always interested in a good conversation.