Jiajing – AI agent development, Python, LLM

An AI-powered evidence intelligence platform for maternal and infant care that helps parents navigate trusted guidance on feeding, sleep, safety, development, postpartum recovery, and baby care. The product organizes fragmented medical and public-health information into structured, searchable pathways and delivers citation-backed answers based on sources such as CDC, WHO, and clinical guidelines. It also acts as the public knowledge layer within a broader ecosystem, supporting connected products for baby tracking and nutrition planning.

• Built and launched a full-stack AI knowledge platform using Next.js, TypeScript, Tailwind CSS, and OpenAI;
• Implemented advanced SEO, AEO, and GEO optimization, including structured data, AI-readable feeds, sitemap routes, metadata strategy, and LLM-friendly endpoints;
• Designed a RAG-based question-answering system that retrieves relevant parenting content and generates source-aware responses;
• Grew the connected mobile app ecosystem to thousands of users;
• Built a social media presence with more than 20K followers;
• Created structured content systems for articles, FAQs, citations, baby food guidance, and topic pages;
• Developed product pages and ecosystem positioning for DearBaby and SolidStart;
• Integrated Supabase for content storage, analytics, waitlist management, feedback, and ingestion workflows.

An AI-native command center for managing autonomous agent workflows that transforms open-ended tasks into structured execution loops with planning, tool usage, verification, and reporting. The product provides visibility and control over agent behavior through trackable runs, decision inspection, and step-by-step execution transparency. It enables reliable task execution by combining structured workflows, governed tool usage, and verifiable outcomes instead of black-box responses.

• Created full-stack architecture for an AI agent workbench using Next.js, React, TypeScript, FastAPI, and Supabase;
• Built backend APIs for creating, managing, tracking, and reviewing autonomous agent runs;
• Implemented an agent execution loop covering planning, tool execution, verification, fallback handling, and final reporting;
• Designed a decision policy layer to classify task types, select next actions, define allowed tools, and enforce completion requirements;
• Built a tool registry supporting repository operations, file search, patching, testing, commits, pull requests, and notifications;
• Implemented structured agent state tracking with run metadata, execution events, token usage, failure records, and completion evidence;
• Added database migrations for agent runs, events, task metadata, user integrations, failures, and task preferences;
• Integrated Supabase authentication, JWT-based API access, and user-specific data storage;
• Implemented worker-based background execution for queued agent tasks;
• Added failure recovery logic with fallback strategies, retry behavior, and verification checks;
• Developed support for coding-agent workflows, including repository cloning, file reading, search, patching, testing, commits, pushes, and PR creation;
• Created risk-aware tool metadata for permission control, approval handling, retry policy, and governance;
• Built and documented API endpoints for tasks, journals, reflections, timelines, tags, smart-agent actions, and runs;
• Prepared deployment configuration for Render, Vercel, and Supabase-backed environments;
• Wrote technical documentation covering agent architecture, API behavior, execution flow, testing, deployment, and implementation details;
• Debugged and resolved web compatibility issues, including Supabase storage behavior, OAuth redirects, and Expo web support;
• Established a scalable foundation for multi-agent collaboration, tool expansion, and external integrations.

An AI-powered journaling and productivity platform designed for startup founders to capture daily thoughts, track milestones, manage tasks, and reflect on progress. The product consolidates fragmented information—ideas, decisions, tasks, and lessons—into a structured workspace with timelines, tagging, and AI-assisted insights. It supports founders in turning raw experiences into organized knowledge and actionable next steps through reflection analysis, task planning, and guided workflows.

• Created full-stack architecture with mobile, web, backend API, database, and deployment support;
• Built journaling functionality for recording founder thoughts, moods, milestones, challenges, lessons learned, and key moments;
• Implemented task management features for creating, prioritizing, tracking, and updating startup-related tasks;
• Integrated AI-powered workflows for task planning, reflection analysis, moment analysis, and brainstorming support;
• Designed and implemented FastAPI backend endpoints for users, journals, tasks, reflections, timelines, tags, task parsing, and smart-agent actions;
• Integrated Supabase for authentication, JWT-based API access, user profiles, database storage, and cross-platform synchronization;
• Added daily reflection workflows to summarize progress, extract lessons, and generate improvement suggestions;
• Built timeline functionality to organize startup events, decisions, highlights, and setbacks;
• Implemented tagging and categorization for organizing entries by themes such as product, team, fundraising, technical work, and growth;
• Developed database migration scripts for journals, tasks, tags, reflections, task metadata, language settings, preferences, and integrations;
• Created Expo-based mobile and web compatibility support, including platform-specific authentication redirects and storage handling;
• Built a Next.js web frontend foundation using React, TypeScript, Tailwind CSS, Supabase, and reusable API utilities;
• Prepared deployment configuration for Render, Vercel, and Supabase-backed environments;
• Wrote technical documentation covering API design, implementation, project structure, deployment, testing, and architecture;
• Debugged and resolved application issues related to authentication, database consistency, task metadata, and cross-platform behavior;
• Established a scalable foundation for features such as story sharing, community interaction, analytics, and AI-driven insights.

A public healthcare directory that helps patients find walk-in clinics, family doctors, and clinics accepting new patients through a searchable interface, map-based discovery, and structured clinic profiles. The product consolidates fragmented healthcare information into a normalized, continuously updated dataset, enabling users to compare clinics and make faster, more informed decisions. It also includes a data management layer with agent-assisted workflows for ingesting clinic data, detecting status changes, and maintaining verified, high-quality records over time.

• Built a searchable Alberta clinic directory with list and map views;
• Added city pages for high-intent searches such as Calgary, Edmonton, and Red Deer;
• Integrated Supabase/PostGIS for clinic storage and location-aware data;
• Imported and normalized hundreds of Alberta clinic records;
• Added clinic profile pages with hours, phone, address, website, map, reviews, and claim flows;
• Implemented SEO support through metadata, sitemap, Open Graph, and Schema.org structured data;
• Created agent APIs for clinic status ingestion, verification, review, and publishing.