Yilmar – Python, React, PostgreSQL

An intelligent dynamic reporting platform that automates project governance and progress tracking across multiple organizational initiatives. The system leverages an autonomous SQL agent to query, aggregate, and analyze multi-source project metrics, generating customizable data visualizations, structured tables, and executive summaries that users can refine via interactive chat before exporting to production-grade PDF and PowerPoint decks.

• Architected and implemented the end-to-end full-stack platform using FastAPI, Python, React, and TypeScript, delivering a high-throughput reporting engine.
• Designed a LangChain-powered SQL agent capable of dynamically inspecting schema definitions, synthesizing optimized read queries, and aggregating project metrics with strict read-only execution sandboxing.
• Built dynamic chart- and table-generation pipelines, allowing users to customize, edit, or regenerate specific report sections via natural language commands and real-time UI controls.
• Engineered backend export workers capable of rendering complex document templates into high-fidelity PDF reports and interactive PowerPoint presentations asynchronously.
• Implemented structured error handling, schema validation using Pydantic, and comprehensive API caching to maintain sub-second response times across heavy analytical queries.
• Containerized all application services using Docker and configured CI/CD delivery pipelines via GitHub Actions targeting AWS infrastructure (ECS Fargate, Aurora PostgreSQL, S3).

A mission-critical mobile application and conversational AI assistant designed for field promoters and community coordinators. The solution resolves operational data bottlenecks by providing an intuitive, voice- and text-enabled Retrieval-Augmented Generation (RAG) interface, enabling field staff with diverse digital literacy levels to instantly query demographic records, family profiles, and assigned social welfare benefits in real time.

• Engineered the cross-platform mobile client with React Native and TypeScript, prioritizing accessibility, minimal network latency, and offline-resilient local caching.
• Architected a robust RAG backend in FastAPI utilizing PostgreSQL/PGVector for semantic search over structured demographic data, sensitive family records, and program guidelines.
• Implemented strict role-based data isolation and HIPAA/data-privacy safeguards, ensuring context retrieval is strictly constrained to authorized field personnel.
• Optimized vector similarity search latency by applying hybrid search strategies (dense embeddings + sparse keyword matching) and pre-filtering metadata before vector distance calculations.
• Built standardized REST endpoints for conversational streaming responses, managing token utilization and session history across asynchronous background workers.
• Deployed containerized backend services to AWS ECS Fargate, configuring automated horizontal scaling policies based on operational peak hours.

A comprehensive document management and real-time collaboration module integrated into an enterprise CRM/CHT ecosystem. The system streamlines end-to-end document lifecycles for heterogeneous file types (PDF, Office, rich media, and audio), featuring in-browser PDF annotation with drag-and-drop workflows, multi-party electronic signature validation, role-based access control (RBAC), and bi-directional integration with Microsoft SharePoint.

• Led the full-stack engineering of the core document repository module using React, TypeScript, and Python/FastAPI microservices;
• Built an interactive PDF workspace supporting drag-and-drop placement, visual in-line annotations, highlight persistence, and digital signature placement directly in the browser;
• Integrated Microsoft Graph APIs for automated synchronization, permissions management, and seamless check-in/check-out of Office files stored in SharePoint;
• Designed a granular role-based access control (RBAC) matrix governing view, edit, download, and audit-log permissions across collaborative workspaces;
• Implemented secure file streaming, chunked uploads, and asynchronous virus scanning pipelines for large file formats leveraging AWS S3 pre-signed URLs;
• Maintained 99.9% uptime by containerizing microservices with Docker and integrating structured audit logs for compliance tracking.

A flexible dynamic form engine and data collection platform capable of creating, versioning, and deploying complex multi-page surveys across various CRM modules. The engine supports over 25 customizable question types — including matrix grids, voice recordings, geolocation, conditional branching logic, and electronic signatures — handling high-variability submission payloads with rapid retrieval performance.

• Designed the full-stack architecture with React/TypeScript for the interactive drag-and-drop builder and Python/FastAPI for execution and submission processing;
• Selected and implemented MongoDB for form definition and submission storage to effectively accommodate extreme schema polymorphism, deeply nested survey structures, and evolving field definitions without schema migration bottlenecks;
• Built dynamic client-side and server-side validation engines that evaluate runtime conditional logic, dependent question visibility, and multi-step validation rules;
• Engineered media capture and storage handlers for rich input types (voice recordings, canvas-based signature vectors, file attachments) uploaded directly to object storage;
• Implemented high-performance aggregation pipelines in MongoDB to enable real-time survey response analytics, filtering, and automated CSV/XLSX export routines;
• Conducted load testing and optimized query indexes to ensure seamless data ingestion during high-concurrency survey rollout campaigns.