Michael Le
Original ↗

> Projects / index

Three systems,
drawn out.

Areas I build in. Each entry opens into a full case-study layout with its diagram.

03 entries

Agents / ConceptCase 01 of 03

Agent Research Assistant

A research assistant concept powered by LLM agents, designed for deeper, structured exploration.

01 PLAN 02 GATHER 03 READ 04 SUMMARISE ↺ HUMAN REVIEW — EVERY STEP INSPECTABLE FIG.01 — AGENT PATHWAY

Problem space

Research questions usually need many sources, steps and checks — more than a single prompt can hold.

Approach

Break the task into agent steps — plan, gather, read, summarise — keeping each step inspectable.

Focus areas

Tool use, structured outputs and keeping a human in the loop.

Edtech / PlatformCase 02 of 03

EdTech Learning Platform

Tools for personalised learning experiences with AI-driven content recommendations.

01 LEARNER 02 CONTENT MAP 03 LANGUAGE MODEL 04 NEXT STEP ↺ LEARNER MOVES ON — MAP UPDATES FIG.02 — LEARNING LOOP

Problem space

Learners move at different speeds and need material that meets them where they are.

Approach

Use language models to organise content and suggest what to study next.

Focus areas

Learner experience, content structure and sustainable growth.

OCR / Work areaCase 03 of 03

Multilingual Document OCR

The area of my work at GMO-Z.com RUNSYSTEM: reading documents across scripts and layouts.

01 SCAN 02 LAYOUT 03 RECOGNISE 04 STRUCTURE 05 DATA FIG.03 — DOCUMENT → DATA

Problem space

Real documents mix languages, scripts, tables and noisy scans.

Approach

Combine recognition with layout understanding to extract structure from unstructured content.

Focus areas

Multilingual text, document layout and turning output into usable data.