Frame
Turn vague stakeholder pain into a decision question, success metric, and caveat list.
Hermes Agent · data analysis operating system
I use Hermes Agent as an AI automation layer across the full data-analysis lifecycle: scoping, data QA, SQL/Python analysis, evidence validation, dashboard design, documentation, GitHub PRs, and Vercel deployment.
Modern data teams win by shortening the path between stakeholder question and trusted decision artifact. My portfolio shows how Hermes Agent makes each stage faster without removing human judgment.
Turn vague stakeholder pain into a decision question, success metric, and caveat list.
Load public data, define schema, inspect fields, document missing operational signals.
Use Hermes to generate data tests, anomaly checks, null checks, and reproducible audit notes.
Build SQL/Python metrics, compare baselines, segment risk, and keep assumptions explicit.
Stress-test interpretation: association vs causality, proxy vs direct proof, signal vs action.
Convert analysis into recruiter-readable dashboard, case narrative, and proof trail.
Push GitHub PR, preview on Vercel, merge to production domain, document next iteration.
Each case demonstrates different analytics muscle: forecasting/planning logic and operations-risk triage.
Forecast baselines become planner-review candidates using WAPE, MAE, Bias %, Forecast Score, FVA framing, ABC/XYZ segmentation, and replenishment assumptions.
Late-delivery exposure, review-score association, and seller/category/region investigation candidates framed without unsupported root-cause claims.
Automation stack
Hermes Agent does not replace analyst judgment. It orchestrates repetitive work: code generation, file ops, browser checks, GitHub PRs, site deployment, and documentation discipline.
Task-executing AI assistant for analysis planning, code iteration, documentation, browser QA, GitHub workflows, and deployment automation.
Generates recruiter-facing layout artifacts and case-study pages, then hands accepted HTML into a publish pipeline.
Every published change has a branch, commit message, reviewable diff, and deployment preview.
PR previews for review, main branch deploys to im-khang.com.
Recruiter readout
I can operate like a lean AI-augmented analytics team: discover the business question, build the evidence model, document limitations, create a decision artifact, and ship it publicly with a modern PR/deploy workflow.