General Assembly has been teaching people to break into tech since 2011, and the company’s AI bootcamp now sits at the centre of what it sells. The pitch is straightforward: take someone who can already write a bit of code and, over a few months, get them to the point where they can build, train, evaluate and ship machine learning systems.
That’s an ambitious promise. It’s also a crowded market, with free courses, cheap subscriptions and rival bootcamps all competing for the same students. So what does the General Assembly AI Bootcamp actually give you for the money? Here’s a clear-eyed look.
What the Program Actually Looks Like
Delivery is live and online for most cohorts, with in-person sessions still running at campuses in cities including New York, London and Sydney. Every cohort gets an instructor, teaching assistants, a dedicated Slack workspace and a schedule that doesn’t leave much room for coasting.
Full-time versus part-time
The full-time immersive runs about 12 weeks at 40-plus hours a week. This is the version people quit jobs for. Part-time tracks stretch similar material across roughly 24 weeks of evening and weekend classes, usually 10 to 15 hours a week once you count homework.
The full-time route suits career changers who can take a financial hit and want to move fast. Part-time suits people staying employed while they retrain, though momentum is harder to maintain when you’re coding at 9pm after a workday.
What you need before day one
No machine learning background is expected. Basic programming comfort is. The admissions process involves a short aptitude assessment, a conversation with an admissions staffer and a pre-work package of roughly 30 to 40 hours covering Python fundamentals. Skip that pre-work and you’ll feel it by week two.
Inside the Curriculum
Curricula get revised as the field moves, and GA has leaned heavily into generative AI since 2023. A typical sequence covers:
- Python and data wrangling — NumPy, pandas, pulling from APIs, cleaning messy real-world datasets
- SQL and data modelling — joins, window functions, querying warehouses
- Statistics and experimentation — distributions, hypothesis testing, A/B test design and the ways results get misread
- Classical machine learning — scikit-learn, regression, classification, gradient boosting, feature engineering, cross-validation
- Deep learning — neural networks in PyTorch or TensorFlow, CNNs for images, transfer learning
- NLP and transformers — tokenisation, embeddings, attention, fine-tuning pretrained models
- Generative AI and LLMs — prompt engineering, retrieval-augmented generation, vector databases, evaluating model output, guardrails
- Deployment and MLOps — packaging models behind APIs, Docker, cloud deployment, monitoring for drift
- Ethics and responsible AI — bias, privacy, model documentation
You finish with four or five portfolio projects plus a capstone. The capstones that land best in interviews use data the student genuinely cares about — Premier League match stats, local housing records, a scraped dataset from their old industry — rather than another Titanic survival model.
Tuition, Financing and the Hidden Costs
Full-time AI bootcamp tuition has hovered around the $16,000 mark in the US market, with part-time tracks several thousand dollars cheaper and international pricing adjusted per city. Fees change, so treat any figure you read, including this one, as a starting point and confirm it with admissions directly.
Funding routes worth asking about:
- Monthly instalment plans, often interest-free over the length of the course
- Scholarships, including ones aimed at women in tech, veterans and underrepresented groups
- Employer sponsorship, which is more common than people assume for students already inside a company
- Third-party loans through GA’s lending partners
GA wound down its income share agreements in most markets a few years back, so if an ISA is what makes the numbers work for you, check whether one is still offered where you live.
Then add the costs nobody puts on the landing page. Twelve weeks without a salary is the big one. Cloud credits for deployment projects, a machine that can handle training runs, and a few hundred dollars for hosting and a domain are the small ones.
What the Day-to-Day Feels Like
Cohorts usually land somewhere between 20 and 30 students, with one lead instructor and a couple of TAs. Mornings tend to be lecture and live coding, afternoons are labs and pair programming, and most days open with a standup where you say what you’re stuck on. Sessions are recorded, which matters when a concept doesn’t land the first time.
The pace is genuinely fast. Budget four to six hours of homework on top of class time, and expect a stretch around week five where everything feels impossible. Career support runs alongside the technical work: portfolio reviews, resume surgery, mock interviews and employer showcase events where capstone projects get presented to hiring partners.
Job Outcomes, Without the Hype
GA publishes outcomes reports by program and campus. Read them properly. Three numbers matter: the response rate, the share employed in a field-relevant role, and the timeframe. Six months is the industry standard.
Typical destinations for graduates look like this:
- Data analyst or analytics engineer, the most common first role
- Data scientist at a mid-size or enterprise company
- Machine learning or AI engineer, usually where the student already had software experience
- AI-adjacent product, solutions or operations roles
Realistic starting salaries in the US sit somewhere between $75,000 and $115,000 in larger markets, lower elsewhere. Graduates without prior technical work often start in analyst roles and move toward machine learning after a year or two. Anyone promising a $200,000 AI engineer job straight out of a 12-week course is selling something.
How It Compares to Other Routes
Free and cheap self-study. DeepLearning.AI, fast.ai, Coursera specialisations and Kaggle competitions cost almost nothing and cover plenty of the same ground. What they don’t give you is a deadline, a person marking your work or a cohort expecting you to show up.
Subscription platforms. DataCamp and similar services run a few hundred dollars a year and are excellent for specific gaps rather than full career transitions.
Other bootcamps. Flatiron School, Springboard, Le Wagon, BrainStation and Codeup all compete here, with different price points, lengths and job-guarantee terms.
Master’s degrees. A two-year program covers the maths and research foundations a 12-week course can’t. It costs far more and takes far longer, but for research scientist roles it’s the honest answer.
Where General Assembly tends to win is structure, human feedback and a career team chasing you about your LinkedIn profile. Where it loses is price and depth.
Who Should Probably Look Elsewhere
Be honest with yourself on a few points. If you can’t commit the hours, the material will pile up faster than you can clear it. If your goal is a research scientist position at a frontier lab, you need a graduate degree, not a bootcamp. If you’re a self-directed learner who finishes what you start, free resources plus three serious portfolio projects will get you a long way for a fraction of the cost.
Visa sponsorship is another one to check early. A bootcamp certificate doesn’t confer work authorization anywhere, and some employers filter on degrees regardless of your portfolio.
Questions to Ask Admissions Before You Pay
Get specifics in writing rather than accepting the marketing line. Ask what the average cohort size has been over the last three intakes, what the instructors’ current industry roles are, and whether you can speak to two or three recent graduates — then actually call them.
Push on outcomes data too: what percentage of the last 12 months of graduates were employed in a field-relevant role within six months, and how many responded to the survey. Ask about the withdrawal and refund policy, how missed weeks are handled if life intervenes, whether cloud credits are included, and how long you keep access to recordings and the alumni network.
Then do one more thing before you commit. Spend a weekend building something small with a free LLM API and a public dataset. If that weekend leaves you wanting more, a structured program will accelerate you. If it leaves you bored, $16,000 is an expensive way to find out.