Every brief in the programme, published up front
48 project briefs across 12 domains. Read the work before you commit four weeks to it — including what each one is assessed against.
AI Agents & Automation
Build autonomous agents that use tools, not chatbots that answer questions.
- 1
Tool-Using Research Agent
Build an agent that answers a research question by calling a search tool and citing its sources.
~10h estimated 2 deliverablesRead the brief → - 2
Multi-Agent Workflow with LangGraph
Split a task across specialised agents that hand work to each other.
~12h estimated 3 deliverablesRead the brief → - 3
Build Your Own MCP Server
Expose a real capability to an AI client through the Model Context Protocol.
~12h estimated 2 deliverablesRead the brief → - 4
Agent Reliability Harness
Measure whether your agent actually works, instead of assuming it does.
~10h estimated 2 deliverablesRead the brief →
Generative AI & LLMs
Go past prompt tinkering into retrieval, evaluation and cost control.
- 1
Document Q&A with Citations
Answer questions over a document set, with every answer traceable to a source passage.
~12h estimated 3 deliverablesRead the brief → - 2
RAG Evaluation Suite
Prove your RAG system works, with a number you would defend in a review.
~10h estimated 3 deliverablesRead the brief → - 3
LLM Cost Optimisation Study
Cut the cost of an LLM feature without materially hurting quality.
~8h estimated 2 deliverablesRead the brief → - 4
Structured Output Extraction Pipeline
Turn messy unstructured text into validated structured data, reliably.
~10h estimated 2 deliverablesRead the brief →
Machine Learning
Classical ML done properly — including the evaluation everyone skips.
- 1
Credit Scoring Model
Predict an applicant's creditworthiness from historical financial data.
~10h estimated 2 deliverablesRead the brief → - 2
Disease Prediction from Medical Data
Estimate disease likelihood from structured patient data.
~10h estimated 2 deliverablesRead the brief → - 3
Handwritten Character Recognition
Classify handwritten digits or characters with a CNN.
~12h estimated 2 deliverablesRead the brief → - 4
Emotion Recognition from Speech
Classify the emotion in a spoken audio clip.
~14h estimated 2 deliverablesRead the brief →
Data Science
Turn a messy dataset into a decision somebody can act on.
- 1
Real-World Data Cleaning Pipeline
Turn a genuinely messy public dataset into something analysable, reproducibly.
~8h estimated 2 deliverablesRead the brief → - 2
Exploratory Analysis with a Finding
Find something in the data that a stakeholder did not already know.
~10h estimated 2 deliverablesRead the brief → - 3
Interactive Analytics Dashboard
Build a dashboard that answers a specific recurring question.
~12h estimated 2 deliverablesRead the brief → - 4
Time-Series Forecast with Honest Error Bars
Forecast a real series and be truthful about the uncertainty.
~12h estimated 2 deliverablesRead the brief →
Computer Vision
Detection, classification and the messy reality of real images.
- 1
Image Classifier with Transfer Learning
Classify images in a domain of your choosing using a pretrained backbone.
~12h estimated 2 deliverablesRead the brief → - 2
Object Detection on Real Footage
Detect and count objects in video that you did not curate.
~12h estimated 2 deliverablesRead the brief → - 3
OCR Document Extraction
Pull structured fields out of photographed documents.
~12h estimated 2 deliverablesRead the brief → - 4
Edge-Ready Model Optimisation
Make a vision model small and fast enough to run on modest hardware.
~10h estimated 2 deliverablesRead the brief →
Python Development
Write Python that other people can run, read and trust.
- 1
CLI Tool with Real Error Handling
Build a command-line tool that behaves well when things go wrong.
~8h estimated 2 deliverablesRead the brief → - 2
REST API with Tests
Build a small API with a test suite you would trust in CI.
~12h estimated 2 deliverablesRead the brief → - 3
Automation Script with Scheduling
Automate a recurring task and make it safe to run unattended.
~8h estimated 2 deliverablesRead the brief → - 4
Package It and Ship It
Turn one of your earlier tasks into something another person can install and run.
~10h estimated 2 deliverablesRead the brief →
NLP Engineering
Make machines read text well enough to act on it.
- 1
Text Classifier with an Honest Baseline
Classify real text, and prove your model beats the dumb approach.
~10h estimated 2 deliverablesRead the brief → - 2
Structured Extraction from Messy Documents
Pull specific fields out of unstructured text reliably.
~12h estimated 2 deliverablesRead the brief → - 3
Semantic Search over Your Own Corpus
Build search that finds things keyword search misses.
~12h estimated 3 deliverablesRead the brief → - 4
Language Model Evaluation Harness
Build the thing that tells you whether a prompt change made it worse.
~10h estimated 2 deliverablesRead the brief →
Full Stack Development
Ship a web app that survives a real user.
- 1
CRUD Application with a Real Database
Build an app that stores data properly and survives a refresh.
~12h estimated 2 deliverablesRead the brief → - 2
Authentication Done Safely
Add sign-in without inventing your own security.
~10h estimated 3 deliverablesRead the brief → - 3
Responsive, Accessible Interface
Make it work on a phone and with a keyboard.
~10h estimated 3 deliverablesRead the brief → - 4
Deploy with CI and Real Monitoring
Get it live, and know when it breaks.
~8h estimated 3 deliverablesRead the brief →
Cyber Security
Find the holes in your own work before anyone else does.
- 1
Audit a Deliberately Vulnerable Application
Find and document real vulnerabilities on a legal practice target.
~12h estimated 2 deliverablesRead the brief → - 2
Secure Code Review
Read code the way an attacker would.
~10h estimated 2 deliverablesRead the brief → - 3
Harden a Server and Prove It
Lock down a system you own, then demonstrate the difference.
~10h estimated 2 deliverablesRead the brief → - 4
Build a Security Tool
Automate one piece of security work you did by hand.
~10h estimated 3 deliverablesRead the brief →
Cloud & DevOps
Make deployment boring, repeatable and reversible.
- 1
Containerise an Application Properly
Write a Dockerfile you would put in production.
~8h estimated 2 deliverablesRead the brief → - 2
CI/CD Pipeline That Blocks Bad Merges
Build a pipeline that catches problems before they ship.
~10h estimated 2 deliverablesRead the brief → - 3
Infrastructure as Code
Define your infrastructure in a file, and rebuild it from nothing.
~12h estimated 2 deliverablesRead the brief → - 4
Monitoring and Alerting
Know your service is broken before a user tells you.
~10h estimated 3 deliverablesRead the brief →
Java Development
Build backend services the way enterprises actually build them.
- 1
Spring Boot REST Service
Build a working API with proper layering.
~12h estimated 2 deliverablesRead the brief → - 2
Persistence with JPA
Model related data and query it efficiently.
~12h estimated 3 deliverablesRead the brief → - 3
Test Suite That Survives Refactoring
Write tests that break when behaviour breaks, not when code moves.
~10h estimated 3 deliverablesRead the brief → - 4
Secure and Package for Deployment
Add authentication and make it deployable.
~10h estimated 3 deliverablesRead the brief →
Business Intelligence
Turn data into a decision somebody actually makes.
- 1
SQL Analysis of a Real Dataset
Answer five business questions with SQL, and defend the answers.
~10h estimated 2 deliverablesRead the brief → - 2
Dashboard Built for a Specific Person
Build a dashboard for a named role, not for everyone.
~12h estimated 3 deliverablesRead the brief → - 3
Metric Definition and Data Quality
Define metrics precisely and find where the data lies to you.
~8h estimated 2 deliverablesRead the brief → - 4
Analysis Presented to a Non-Technical Audience
Communicate a finding to someone who will not read your notebook.
~8h estimated 2 deliverablesRead the brief →
Ready to Transform Your Enterprise with AI?
Book a free 30-minute consultation with our AI architects. Discover how agentic AI can automate your workflows and multiply your team's output.
No commitment required · Free architecture review · NDA protected