MAIYUAN logo

Internship

Algorithm Intern

MAIYUAN

Industrial data analysis, predictive modelling, and AI-assisted workflow design for high-salinity wastewater resource recovery.
Industrial DataSoft SensorPredictive ModellingAI Workflows

Internship Scope

Industrial resource recovery

MAIYUAN develops resource-recovery and zero-liquid-discharge solutions for high-salinity industrial wastewater using advanced evaporation and crystallisation technologies.

Featured Case · 01

Time SeriesFeature EngineeringValidation Design

Water Production Measurement & Soft Sensor

Abstract data signals converging into an estimated water-production curve.
AI-generated conceptual visual; not a real site or equipment record.

The Challenge

Discharge readings did not always reflect real-time water production: storage changes, pump operation, sensor timing and operating transitions all distorted them. Building a reliable production label was part of the modelling problem.

A soft sensor estimates a process value that is difficult to measure continuously from other available operating signals.

My Role

  • Mapped evaporation, condensation, storage and discharge into a measurement model the business could read.
  • Rebuilt minute-level production labels from anonymised discharge and storage changes, and checked them against independent metering.
  • Built features from upstream process signals only, excluding label sources and downstream outcomes to prevent target leakage.
  • Validated in time order and on a later operating regime to test generalisation beyond the training window.

Method Flow

  1. 01Reconstruct the measurement chain
  2. 02Build a reliable minute-level label
  3. 03Model upstream process history
  4. 04Validate forward across operating regimes
  • Measurement reliability before model complexity.
  • Multi-million-row source dataset with more than 200,000 stable operating minutes.
  • Time-ordered evaluation instead of random train/test splitting.

Offline Validation

Measured forward, qualified carefully

0.0409 t/mininternal time-ordered validation MAE
12.48%internal time-ordered validation WMAPE
0.0375 t/minlater operating-regime validation MAE
9.30%later operating-regime validation WMAPE

Offline validation results; the model had not yet entered production deployment.

Measurement assumptions still need field confirmation, and steady-state rules must be refit on training data before a final benchmark.

Featured Case · 02

StoryboardProgrammatic VideoMedia QA

AI-Assisted Video Research & Production

Conceptual storyboard, timeline, audio, and validation layers for an AI-assisted video workflow.
AI-generated conceptual visual; not a real site or equipment record.

The Challenge

Enterprise video needs richer visuals, but generated imagery must never pass as evidence of a customer, site, product capability or operating result.

Production Principles

  • Use verified source material for factual claims; use AI only for low-risk atmosphere or transitions.
  • Keep scripts, storyboards, assets, timelines and render records as separate, reviewable layers.
  • Check duration, encoding, audio, safe areas and file identity before human review.

My Role

  • Reframed tool selection around business scenarios, evidence boundaries and acceptance gates.
  • Designed a governed path from brief and storyboard through asset routing, programmatic rendering, media checks and approval.
  • Built a vertical company-profile review film with full voice and captions, and logged its status so it could not be mistaken for a public release.

Governed Workflow

  1. 01Scenario and evidence brief
  2. 02Storyboard and asset routing
  3. 03Programmatic edit and render
  4. 04Automated checks and human approval

Internal Review Build

30 secondsvertical internal review film
9:161080 × 1920 at 30 fps
13-shotapproved storyboard and scene route
16 automated QA checksvoice, captions, safe areas, and media output

An internal review version has been rendered with voice and captions. Final approval is still pending, so no review media is included here.

Supporting Contributions

Tools and organizational enablement

01

Industrial Trend Explorer

Built and released an offline-capable tool for exploring long-range industrial time series.

  • Filters by process, equipment group and field, keeping valid negative values, cleaning markers and normalised views.
  • Built with Python, Pandas, PyArrow and native Canvas, with no front-end runtime dependencies.
  • Queries approximately 4.009 million rows without loading them all into memory, at a median of about 47 ms; gzip cuts a typical response from 3.18 MB to 0.80 MB.
  • Covered by 15 automated tests for querying, cleaning, sampling, compression and offline use.

02

Executive AI Enablement

Designed a one-hour executive training programme that turned AI capabilities into reviewable management workflows.

  • Built each demo as business scenario → AI operation → deliverable → management value, not a feature tour.
  • Covered documents, slides, spreadsheet analysis, browser research, recurring automation and reusable prompts, with safety boundaries beside each.
  • Turned the preparation into a reusable Skill and SOP for future sessions.