AI internships you can finish

59 self-paced drills. Each one builds a system companies ship today, grounded in the papers the technique came from.

Or start from a goal

Curated paths through the drills above, grouped by what you're trying to become.

Goal

I'm a leader and I want to evaluate and govern AI investments in my organization

📚Basic → Advanced

10 drills

Goal

I'm a security professional and I want to test AI/ML systems for vulnerabilities

📚Basic → Advanced

3 drills

Goal

I'm a business professional and I want to use Claude as an AI copilot in my day-to-day work

📚Pick your function

3 drills

Agent Blast Radiusadvanced

Tell an agent that answers apart from one that acts, and explain why "automation" hides that difference

5 sections

Automated Money Managementbeginner

Extract structured data from a receipt photo and know when to double-check the extraction

5 sections

Build a Codebase Knowledge Graphadvanced

Parse a Python repo into a navigable call/import graph using only the stdlib ast module

9 sections · Sourcegraph, GitHub

Build a Disciplined Coding Agentadvanced

Build a plan-gate and a critique-gate for a coding agent, and wire them into one toggleable loop

10 sections · Anthropic, Cognition, Cursor

Build a Model-Serving Engineadvanced

Build a model-serving engine from scratch — the prediction API, batching queue, and cache behind a production inference service

10 sections · Netflix, Uber, NVIDIA

Build an Agent Memory Engineintermediate

Build long-term memory for AI agents

10 sections · Maximem, Mem0, Zep

Build a ReAct Agentintermediate

Build the ReAct loop from scratch in pure Python — reason, act, observe, repeat

9 sections · OpenAI, Anthropic, LangChain

Build a Recommender from Scratchadvanced

Build a recommender from scratch — similarity, collaborative filtering, and matrix factorization over a real ratings dataset

13 sections · Netflix, Amazon, Spotify

Build a Two-Tower Retrieval Engineadvanced

Build a two-tower retrieval model — separate user and item encoders trained into one shared embedding space

11 sections · Netflix, YouTube, Spotify

Build GPT from Scratchadvanced

Build a decoder-only GPT language model from scratch in PyTorch and generate text with it

10 sections · OpenAI, Google DeepMind, Anthropic

Building a Coding Agentadvanced

Build a coding agent that reads, edits, and runs code in a tool-use loop

13 sections · Anthropic, Cursor, GitHub

Building a Customer-Segmentation Enginebeginner

Build a customer-segmentation engine that turns a table of customers into named, actionable segments

11 sections · Spotify, Amazon, Airbnb

Building a House-Price Predictorbeginner

Build a house-price predictor that estimates a home's value from its features

10 sections · Zillow, Redfin, Opendoor

Building an AI Email Campaign Management Agentadvanced

Build an autonomous AI agent that plans, writes, schedules, and sends email campaigns

13 sections · Google, Mailchimp, Klaviyo

Build vs. Buy vs. Promptadvanced

Reframe a vague "we want AI to do X" ask into a real problem statement

5 sections

Build Your AI Stackbeginner

Map your work onto the six jobs AI is actually good at

7 sections

Build Your First AI Servicebeginner

Build and ship a working AI service end to end — train a model, serve it over HTTP, protect it, and containerize it

10 sections · Hugging Face, OpenAI, Docker

CoWork: AI for Sales and Marketingadvanced

Run account and market research with an AI assistant

25 sections · Anthropic, Salesforce, HubSpot

CoWork for the Chief of Staffadvanced

Draft executive communications and weekly summaries with AI

6 sections

CoWork for the Clinicadvanced

Draft referral letters and patient messages with AI

6 sections

CoWork for Your Shopbeginner

Set up Claude CoWork with a shop memory file so it knows your business every session

5 sections

Hallucination & Guardrailsadvanced

Name the specific reason a model hallucinated — missing grounding, missing abstention pressure, or both

6 sections

Language Modeling from First Principlesbeginner

Build a predictive-text engine from scratch — the next-word suggestions behind a phone keyboard

8 sections · Google, Apple, Microsoft

Learn Fast Auto Encodersintermediate

Denoise images with neural nets

4 sections · PayPal, Feedzai, Topaz Labs

Learn Fast: Autonomous Agents on FrozenLakeintermediate

Build a self-learning agent

4 sections · DeepMind, OpenAI, Amazon

Learn Fast: Build a Real-Estate Data Appintermediate

Build a multi-page Streamlit data app

5 sections

Learn Fast Convolutional Neural Networksbeginner

Build an image classifier

5 sections · Google, Apple, Tesla

Learn Fast Deep Q Learningintermediate

Build a Deep Q-Network

6 sections · DeepMind, Google

Learn Fast Develop Emotion Support Agentintermediate

Build an emotion-support chatbot

4 sections · Replika, Character.AI, Inflection Pi

Learn Fast: Docker for Deep Learningintermediate

Containerize an ML project with Docker

4 sections · AWS, Google, NVIDIA

Learn Fast Dynamic Pricing using Reinforcement Learningadvanced

Build a dynamic-pricing engine

4 sections · Alibaba, Uber, Lyft

Learn Fast Generative Adversarial Networkadvanced

Generate images with GANs

4 sections · NVIDIA, Adobe

Learn Fast LangGraphintermediate

Build stateful agents with LangGraph

4 sections · Klarna, Uber, J.P. Morgan

Learn Fast LLM Applications using LangChainintermediate

Build LLM apps with LangChain

8 sections · Klarna, Rakuten, Replit

Learn Fast Neural Network Quantizationintermediate

Shrink models with quantization

7 sections · NVIDIA, Qualcomm, Apple

Learn Fast Policy gradient RL Algorithmsadvanced

Train policy-gradient agents

4 sections · OpenAI, Anthropic, Google

Learn Fast Q Learningbeginner

Build a Q-learning agent

7 sections · AWS, Google, Microsoft

Learn Fast SQL Agentintermediate

Query a database in plain English

4 sections · Snowflake, Databricks, Google

Learn Fast: Structuring a Deep Learning Projectintermediate

Lay out a clean, scalable deep-learning project

3 sections · DrivenData, Docker

Learn Fast: The Deep Learning Project Life Cycleintermediate

Scope a deep-learning project end to end

4 sections · Google, AWS, Tesla

Learning Assistantintermediate

Build a multi-tab Streamlit app

6 sections · OpenAI, Streamlit, Groq

Legal-Clause GPT — Fine-Tune GPT-2 on Contractsintermediate

Fine-tune GPT-2 on real legal contract clauses so it tags a clause with its type

9 sections · Harvey, Casetext, LexisNexis

LLM powered Agentsintermediate

Build an LLM-powered agent

5 sections · Intercom, Zendesk, GitHub

LoRA from Scratch — Parameter-Efficient Fine-Tuningadvanced

Build a LoRA adapter from scratch — freeze a pretrained weight and learn a low-rank update in its place

9 sections · Hugging Face, Microsoft, Databricks

ML, GenAI, or Just Software?advanced

Sort a client request into classical ML, GenAI, or plain software — and decompose it when it's more than one

6 sections

Neural Collaborative Filteringadvanced

Build GMF, MLP and NeuMF in PyTorch and train them on real MovieLens interactions

13 sections · YouTube, Meta, Pinterest

Neural Network for CyberSecurityintermediate

Detect DDoS attacks with ML

4 sections · Cloudflare, AWS, Akamai

Phishing & Malicious-URL Detectorintermediate

Build a classifier that flags phishing and malicious URLs from the link text alone

10 sections · Google Safe Browsing, Microsoft Defender, Proofpoint

Prompt Injection & Data Leakageadvanced

Tell direct prompt injection apart from indirect, and explain why retrieved content is untrusted input

5 sections

Reading a Vendor Demoadvanced

Recognise a demo as a cherry-picked trace, not evidence of production behaviour

5 sections

Set Up Your Shop's Chatbotbeginner

Tell a live, auto-replying chatbot apart from a draft-and-review assistant, and why that changes the risk

7 sections

Ship and Monetize an AI Agentadvanced

Package a Python AI tool into a pip-installable distribution ready for PyPI

11 sections · Cursor, Tabnine, Sentry

Speed Marketing for Your Shopbeginner

Draft a month of promotional copy — flyers, sale copy, newsletters — from your shop's real details

5 sections

The Data Readiness Auditadvanced

Audit a client's data against six readiness criteria before any model work starts

6 sections

The Latency Reality Checkadvanced

Explain why prefill is fast and decode is slow, and what that means for any "sub-200ms" claim

6 sections

TLS / Encryption Config Auditorbeginner

Build a TLS auditor that grades a server's encryption the way SSL Labs does

8 sections · Qualys, Cloudflare, Let's Encrypt

Token Economics & Cost Modelingadvanced

Turn a vendor's per-token price into a real monthly cost, not a naive multiplication

6 sections

Vision Transformersintermediate

Build a Vision Transformer

5 sections · Google, Meta, Hugging Face

What Can I Actually Automate?beginner

Sort any shop task into Safe / Check / Never by what happens if the automation is wrong

5 sections

AI internships you can finish

Or start from a goal

I'm a software engineer and I want to learn agentic programming

I want to become an ML engineer who ships and operates models in production

I'm a software engineer and I want to specialize in computer vision

I'm a developer and I want to fine-tune small LLMs for my own use case

I'm a leader and I want to evaluate and govern AI investments in my organization

I'm a security professional and I want to test AI/ML systems for vulnerabilities

I'm a business professional and I want to use Claude as an AI copilot in my day-to-day work