AI Systems Engineering
6 Weeks
·Cohort-based Course
The class where you go from Hugging Face Transformers to building and deploying a full inference stack that real companies pay millions for
AI Systems Engineering
6 Weeks
·Cohort-based Course
The class where you go from Hugging Face Transformers to building and deploying a full inference stack that real companies pay millions for
Author of LLMOps, GPU Engineering Books
Course overview
From AI Engineer to AI Systems Engineer
If you’re an AI engineer who can already prompt or fine-tune models but you’ve never been able to answer questions like:
- Why is my 70B model using 120 GB of VRAM and still slow?
- How do I serve 500 concurrent users on 4×H100s without going broke?
- What actually happens inside FlashAttention / PagedAttention / tensor parallelism?
- How do I make money (or save my company millions) running open models in production?
… then this is the course you’ve been waiting for.
In 6 intensive weeks you will build, profile, optimize, distribute, and ship a complete LLM inference system that can profitably serve LLaMA-3-70B, Mixtral, DeepSeek, or any 70B–405B-class model, including stateful, tool-calling workloads.
You’ll leave with:
- A live, public (or internal) inference API that beats most commercial providers on price/performance
- Hard numbers you can quote in interviews or to your boss
- A portfolio project that gets you hired at the top inference companies (Fireworks, Together, Groq, OctoAI, etc.)
- No prior CUDA. Just 4×2-hour live sessions.
Who is this course for
01
This course is for AI engineers, ML infrastructure engineers, and backend developers who want systems-level mastery
02
AI/ML engineers who know PyTorch/Transformers but feel stuck at the research to production gap
03
Founders building LLM apps who are tired of burning money on OpenAI / Engineers who own inference cost and need to 10–50× it
Prerequisites
You have shipped at least one non-trivial Python project
You can already load and run an LLM using HF Transformers
Comfortable with basic terminal/SSH, git, yaml & json files
What you’ll get out of this course
A live, production-grade LLM inference service running 70B+ (or MoE) models
Projects and hard metrics that you can quote in interviews, share with your team:
- I served LLaMA-3-70B at 108 tokens/s and $0.32 per million output tokens
- My system handled 500+ concurrent stateful, tool-calling requests with <900 ms first-token latency
Optimize LLM inference across the full stack
From kernel-level tuning to distributed execution, you’ll learn how to speed up every layer of the LLM serving pipeline. This includes batching and caching strategies, load balancing, inference engine configuration, and parallelism techniques (data, tensor, pipeline, and expert).
A permanent LLM Cost Calculator
Being able to predict real-world cost and throughput for any model/hardware/parallelism combo within ~10 % accuracy
Battle-tested profiling skills
You'll be able to take any slow inference setup, find the real bottleneck in <15 minutes, and fix it
Deep, intuitive understanding of key concepts
You would be able to explain prefill vs decode, KV cache, tensor/pipeline/sequence parallelism, PagedAttention, FlashAttention-2 and when each actually matters
Confidence to choose the right inference engine
Learning how to choose between vLLM, TensorRT-LLM, DeepSpeed-FastGen, TGI and hardware (H100 vs B200 vs 4090 vs Groq) before spending a single dollar
What’s included
Live sessions
Learn directly from Abi Aryan in a real-time, interactive format.
Guest Speakers
Learn from industry professionals and their experiences.
Lifetime access
Go back to course content and recordings whenever you need to.
Community of peers
Stay accountable and share insights with like-minded professionals.
Certificate of completion
Share your new skills with your employer or on LinkedIn.
Maven Guarantee
This course is backed by the Maven Guarantee. Students are eligible for a full refund up until the halfway point of the course.
Course syllabus
6 live sessions • 18 lessons • 5 projects
Week 1
Foundations of LLM Inference
Jan
17
Session 1
Week 2
GPU-Level Optimization & Profiling
Jan
24
Session 2
Week 3
Parallelism & Distributed Inference
Jan
31
Session 3
Week 4
Building Production-Ready, Stateful Inference Systems
Feb
7
Session 4
Week 5
Guest Lectures from Industry Experts
Feb
14
Session 5
Week 6
Demo Day: Design & Optimize Your Own Scalable LLM Inference Stack
Feb
21
Session 6
Join an upcoming cohort
AI Systems Engineering
Valorants
$1,999
Dates
Payment Deadline
Meet your instructor
Abi Aryan
Abi Aryan is the founder and lead research engineer at Abide AI, a deep tech startup working to integrate causal reasoning in AI systems, where she is developing the decision making stack for agentic applications in enterprise.
She is also the author for an O'Reilly book titled LLMOps and is writing her next book titled GPU Engineering for AI Systems with Packt.
Learning is better with cohorts
Active hands-on learning
This course builds on live workshops and hands-on projects
Interactive and project-based
You’ll be interacting with other learners through breakout rooms and project teams
Learn with a cohort of peers
Join a community of like-minded people who want to learn and grow alongside you
What people are saying
Maria Vechtomova
Frequently Asked Questions
Join an upcoming cohort
AI Systems Engineering
Valorants
$1,999
Dates
Payment Deadline