case study

agentic learning coach

a multi-agent ai system that builds your personalized learning curriculum and guides you day by day

role: ai engineer & full stack devtype: personal projectyear: 2026

langgraphfastapipostgresqlnext.jslangsmithgroqtavily

01 / problem

learning without structure is just consuming content

most people trying to learn new tech have no clear roadmap, waste hours searching for resources, and have no way to verify they actually understood what they learned.

existing tools either give generic roadmaps or require too much manual effort to follow through.

no roadmap

wasted hours

no verification

02 / solution

an agent that thinks, plans, and holds you accountable

user describes their goal and available hours, a multi-node langgraph agent searches the web intelligently, generates a day-by-day curriculum, saves it to postgresql, and serves it through a fastapi backend. every day the user gets their task, studies, then gets quizzed. score below 70% triggers a review day automatically.

web search

daily plan

smart quiz

auto review

03 / architecture

how the planner agent works

planner pipeline

extract topics

parses user goal into key concept nodes

search resources

queries web & docs for high-quality content

generate curriculum

sequences lessons chronologically

save to database

commits structured json data to db

daily session pipeline

fetch task

study

quiz

score + save

04 / results

what it actually does

curriculum

42 days

generated curriculum length

accuracy

100%

quiz score on first real test

complexity

4 nodes

in the planner pipeline

the pipeline generates custom dynamic daily lectures loaded with target code snippets and references curated directly from recent documentation.
interactive score validation monitors learning curve spikes, automatically injecting remedial study days where scores slip.

05 / stack