Internal Ops Assistant
A chat-style assistant concept that lets leadership ask questions against trusted budget, vendor, payroll, and employee cost data.
Postgres data layerSQL/Python logicLLM answer layerHuman approval loop
Agent architecture
Agent concepts for analytics, operations, and future AI products.
Agents are framed as controlled operating workflows: context, tools, task state, validation, and human review.
SQL QA Agent
Agent pattern for checking source totals, modeled facts, variance thresholds, and dashboard readiness before reporting.
Source checksVariance rulesException summariesReview routing
Chorus Product Prototype
In-development iPhone and web product for learning AI agent fundamentals, comparing platforms, and saving reusable skills.
PrototypeSwiftUI appSupabase schemaLaunch assets
Agent mode is task architecture, not an uncontrolled chatbot.
The strongest agent pattern in this portfolio is Ghost-AI: user prompt, project context, background task, AI-generated canvas updates, visible status, and reviewable output. Chorus extends that direction into agent education and reusable skills.
step 01
Define task and business objective
step 02
Retrieve trusted context or project state
step 03
Choose tool path: SQL, API, canvas, document, or workflow
step 04
Run task with visible status and intermediate state
step 05
Validate output against source data or product intent
step 06
Route to human review before decision or release
Workflow orchestration
Trigger.dev-backed tasks move long-running AI work out of request handlers and into observable runs.
Validation and review
Agent outputs are treated as drafts until checked against source data, user intent, and product constraints.
Reusable skills
Chorus provides the product direction for reusable agent skills, platform comparison, and learning workflows.