I design and ship production AI/ML systems — credit decisioning platforms, fraud detection, document intelligence, and the data infrastructure underneath them — for fintech and financial-risk environments. Five years in, my work has shifted from building individual models to architecting the systems they live in: orchestration, MLOps, hybrid cloud data platforms, and the engineering decisions that let ML survive contact with production.
I'm currently focused on AI/ML system architecture — designing platforms that are auditable, scalable, and owned end-to-end rather than stitched together from vendor APIs.
Currently: Senior Data Scientist @ Davivienda El Salvador — built the bank's credit pre-approval platform and rebuilt its credit scoring infrastructure into an internally-owned, auditable decisioning system. Previously: Machine Learning Engineer @ Chivo Wallet — sole ML engineer; owned architecture standards, MLOps, and production ML across fraud detection, OCR/NLP, and a hybrid on-prem/GCP data platform.
- Systems over scripts. A model is the easy part. The hard part is the orchestration, monitoring, and failure modes around it — that's where most production ML actually breaks.
- Own the critical path. At Davivienda, that meant replacing a third-party credit scoring vendor with an internal, auditable pipeline. Dependencies you don't control are technical debt with someone else's name on it.
- Build for the reviewer, not just the model. Some of the highest-leverage tools I've shipped weren't predictive models — they were graph-based investigation tools that cut fraud review time from days to minutes by making evidence legible to a human.
Most of my production systems (credit decisioning, fraud platforms, internal data infra) live in private company repositories and can't be shared publicly. The projects below are independent builds — designed to demonstrate the same architectural thinking using public data and from-scratch system design.
| Project | Focus | Status |
|---|---|---|
| 🔜 Credit Decisioning Engine (rebuild, public version) | Multi-stage decisioning pipeline, rules + ML scoring, full audit trail | In progress |
| Autonomous Multi-Agent Orchestrator | LangGraph-based task decomposition, persistent memory, async tool execution, FastAPI streaming backend | Documentation refresh in progress |
| Enterprise RAG Pipeline | Retrieval-augmented generation over unstructured documents, FAISS + evaluation loop to reduce hallucination | Documentation refresh in progress |
(Links above will go live as each repo's architecture docs are finished — see pinned repos for the current state.)
Languages & Core ML
GenAI & Agentic Systems
Data & Platform Engineering
Cloud & MLOps
Backend & Architecture
B.S. Computer Science, Francisco Gavidia University · Azure AI Fundamentals (AI-900) · Google Cloud ML/MLOps/Vertex AI training track
📫 Reach me at christian.valldaresp@gmail.com or on LinkedIn.