7
production AI agent systems delivered
Built for local and international banking environments at Finaipro.
Software engineer building systems, automation, and applied AI where they create real leverage.
I design production software, automate high-friction workflows, and apply AI to research, operations, and customer-facing systems without treating AI as the whole product.
Medellín, ColombiaBackend and full-stack deliveryAgentic systems and workflow automationApplied AI for production and research
system scopesoftware · automation · ai
operating modelobserve · orchestrate · ship
Inputs
Systems
Automation
Delivery
system scopesoftware · automation · ai
01Inputs
02Systems
03Automation
04Delivery
Proof of range
I work across engineering, automation, research, and AI delivery.
The common thread is system leverage: shipping software that removes friction, scales better than manual work, and stays usable in production.
60s → 20s
response time on complex agents
Reduced by reworking cache layers, persistent memory, and learning loops.
100s → 10s USD
AI search cost reduction
Drove down search spend by optimizing RAG and vector database usage.
+20%
payment commitments
Improvement from a voice-enabled negotiation workflow powered by multi-agent reasoning.
Capability lanes
Not an automation-only profile.
I can build the underlying system, automate the workflow around it, and apply AI when it improves the outcome rather than just adding novelty.
Software Systems
I build the application layer around the AI: APIs, services, product surfaces, and the integration points that make systems durable.
- Backend architecture with Python, FastAPI, Spring Boot, and SQL-heavy workflows
- Full-stack delivery with React and Next.js for internal and client-facing products
- Deployment, observability, and integration design across Azure, Docker, and Kubernetes
Automation Systems
I turn operational work into repeatable systems, from internal tools to multi-step automation flows with human review where it matters.
- Role-based workflow automation for enterprise teams
- Agent orchestration with micro-agent patterns and tool calling
- Secure MCP-enabled access to internal architecture, analytics, and execution layers
Applied AI
I use LLMs and retrieval systems as practical infrastructure, choosing the model and system shape that fits the job instead of forcing an AI-first answer.
- RAG, LLMOps, prompt design, and evaluation for production use cases
- Fine-tuning and model adaptation with open-source and commercial models
- Multi-agent systems for negotiation, search, and enterprise decision support
Research & Data Science
I work comfortably in experimentation-heavy environments where the right answer comes from iteration, measurement, and validation rather than intuition alone.
- Transformer-based time-series forecasting research
- Validation frameworks for experimentation and model comparison
- Autonomous market research pipelines that enrich downstream ML systems
Experience
Production delivery backed by research depth.
B.Eng. in Systems and Computation, EIA University (2021 — 2025, CGPA 4.29/5). I move between enterprise delivery and experimental work without losing product or engineering discipline.
February 2026 — Present
AI Engineer · Finaipro
Medellín, Colombia
Production AI engineering across enterprise banking systems, combining orchestration, infrastructure, and delivery discipline.
- Delivered seven production AI agent systems for international and local banks while supporting internal ML products.
- Built a multi-agent orchestration hub where agents can create and manage role-based agents with tool calling, RAG, and human-in-the-loop workflows.
- Implemented observability with Langfuse and handled frequent deployments on AKS and Azure Functions.
- Created MCP servers for secure interaction with analytics and SQL-driven internal systems.
software engineeringautomationapplied ai
August 2025 — January 2026
Data Scientist · Finaipro
Medellín, Colombia
Research-focused work on forecasting and early agentic investment systems for cross-border finance use cases.
- Led transformer-based time-series forecasting research for stock-return prediction with exogenous variables.
- Built the initial experimentation and validation framework for forecasting model evolution.
- Designed early agentic investment assistants for cross-border banking use cases in Spain and France.
researchapplied aisoftware engineering
December 2024 — May 2025
Data Scientist Intern · Sistecrédito
Envigado, Colombia
Applied AI work that combined model adaptation, workflow design, and product delivery for internal teams.
- Fine-tuned LLMs for automated PQRSF classification across more than 50 categories, reaching 80% accuracy.
- Built a RAG-powered multi-agent payment negotiation workflow that increased payment commitments by 20%.
- Developed a WebSocket-connected Next.js demo application so non-technical teams could test and review the system.
applied aiautomationsoftware engineering
Selected systems
Enterprise AI systems and traditional software builds.
The portfolio includes internal enterprise delivery, applied AI workflows, and public software systems to show breadth rather than a single niche.
Multi-agent orchestration hub
A production orchestration layer where enterprise agents can create, coordinate, and manage role-based agents with tools, retrieval, and review loops.
Turned ad hoc agent experiments into a reusable platform for banking delivery.
Visit FinaiproAutonomous market research engine
A persistent research agent that tracks macroeconomic signals for the S&P 500 and STOXX 600, enriching forecasting workflows with Refinitiv data.
Connected ongoing research automation with downstream ML and market-analysis pipelines.
About the teamNegotiation and classification workflows
A combination of LLM fine-tuning, RAG, and voice-enabled multi-agent interactions for collections and service operations.
Reduced manual work, improved classification quality, and increased payment commitments.
Visit SistecréditoShortOMatic
A URL-shortening platform built as a microservices system with authentication, caching, metrics, redirects, and self-hosted deployment.
Proof that my work also includes traditional product and infrastructure engineering beyond AI-specific delivery.
PythonTypeScriptSQLFastAPINext.jsLangChainLangGraphAgnoAzureDockerKubernetesPostgreSQLRedispgvectorPineconeQdrantLangfuse
Contact
If the brief needs solid engineering, automation, or applied AI, I can work across the stack.
Best fit for teams that need systems that actually ship: internal tools, workflow automation, AI-enabled products, or research-backed engineering work.
Build the system first. Automate the friction next. Apply AI where it matters.
That is the operating model behind my work, whether the result is an internal workflow, a production service, or an AI-enabled product surface.
Email me directly