- Stochastic LLMs
- Chain-of-Thought Prompting and Few-shot Prompting
- Prompt Failure Modes
- Structured Outputs
- Prompting Reliably
- Regression Harness
- Prototypes and Demonstrations: 6
- System Design: Fact Checking Agentic System
The Applied AI Masterclass
A hands-on, no-fluff, and highly practical masterclass that helps you become great at designing, implementing, and shipping AI systems at your workplace.
Enrollments for July 2026 Cohort are open.
If you are looking for system design courses, I offer two - one for beginners and one for experienced engineers.
Key Details
- Super hands-on and practical - focused on you building AI systems at your workplace
- Build and explore 24 prototypes for deep, practical understanding
- Design and brainstorm 7 agentic systems combining AI and resilient system design
- LLM credits are not included; you will need your own API keys for prototyping
- Examples use Gemini, but you are free to use any model
- Lifetime access to all session recordings
- All systems will be built in Python
- Course does not cover Machine Learning or Deep Learning, the focus is to understand how to leverage LLMs, understand different patterns, pitfalls, and build great systems using them.
Why Learn Applied AI Now?
- AI is becoming a core layer of modern products
- Building with LLMs, agents, and AI workflows is now a key engineering skill
- Reliable AI systems require memory, tools, retrieval, workflows, and evaluation - not just LLM call
- AI is the biggest opportunity today for engineers who can integrate AI into real-world products
Program outline
6
SESSIONS
SESSIONS
24
PROTOTYPES
PROTOTYPES
7
SYSTEMS
SYSTEMS
Prompt LLMs Reliably
Week 1 - Session 1
Learn how to design reliable prompts, identify and prevent common LLM failures, and evaluate outputs systematically. This session establishes the foundation for building dependable AI applications.
Tool Use and RAG in Production
Week 1 - Session 2
Learn how production AI systems use tools and retrieval reliably at scale. This session covers the foundations of robust RAG systems, improving retrieval quality, reducing hallucinations, and handling real-world operational challenges.
- Tool Use and Tool Schema Design
- Parallel Tool Calls and Partial Failures
- Hybrid Search and Metadata Filtering
- Re-ranking - Cross Encoders and Bi-encoders
- Query Rewriting
- Semantic Caching
- Prototypes and Demonstrations: 7
- System Design: RAG over 10M docs, no Hallucinations
Building Agents That Work
Week 2 - Session 1
Learn the core patterns behind AI agents, how they reason, plan, and act across complex tasks. This session is essential for building reliable agents that can operate beyond single LLM calls.
- Observe-Think-Act Loop
- Ralph Loop
- ReAct Loop
- Plan and Execute Loop
- File System as Context
- Long Running Agents: Checkpoint and Resume
- Human in the Loop
- Prototypes and Demonstrations: 7
- System Design: Paper to Code Agent
Multi-Agent Systems and Memory
Week 2 - Session 2
Learn how multiple AI agents collaborate, share memory, and coordinate effectively in complex systems. This session is essential for building scalable, maintainable agent architectures that remain reliable and observable in production.
- Memory Architecture
- Working Memory Management and Budgeting
- Memory Summarization and Write Strategies
- Orchestrator + Specialist Pattern
- Critic + Refiner Pattern
- Mixture of Agents Pattern
- Deadlocks in Multi-agent Systems
- Prototypes and Demonstrations: 5
- System Design: AI Incident Auto-remediation System
Evaluation, Safety, and Agentic Systems
Week 3 - Session 1
Learn how to evaluate AI systems systematically, measure performance, and identify weaknesses before deployment. This session is critical for building trustworthy, safe, and continuously improving AI applications in production.
- What makes a good eval?
- LLM-as-Judge
- Non-negotiable Human Evals
- Exploratory and Adversarial Evals
- Prompt Injection
- Prototypes and Demonstrations: TBD
- System Design: AI Code Reviewer System
- System Design: Self-Updating AI Documentation Platform
Production Architecture and Scale
Week 3 - Session 2
Learn the architectural patterns behind scalable AI systems, from reliability and cost management to deployment and observability. This session is essential for building production-ready applications that perform consistently at scale.
- Prompt Caching
- Streaming and Partial Results
- Canaries and Shadow Deployments
- Cost Attribution
- System Design: Natural Language Workflow Engine
- Brianstorming Production Pitfalls
Program Pre-requisites
You have some basic familiarity with Python
You have gone through, read, and understood the Pre-reads of this course
You have 3 weekends of time that you can dedicate towards upskilling
You have a Gemini or OpenAI or Anthropic API access
July 2026 Cohort
3 weeks • 4th July, 2026
8:00 pm to 11:00 pm IST on Saturdays and Sundays
During the live sessions, you will
- learn the intuition behind designing, building, and implementing AI systems
- brainstorm, interact, and learn from the entire cohort and their experiences
- build prototypes & understand implementation details and operational challenges
Enroll Now
Early bird discount is now available and it ends on 20th June at 11:59:59 pm IST
₹60,000 ₹50,000 $650
inclusive of all the taxes
YOU'LL GET
✔ Live Classes on Weekends IST
✔ Lifetime access to the cohort recordings
✔ Lifetime access to the Network and Community
✔ Open forums and interaction with the cohort
✔ Doubt resolution during and post live sessions
✔ 2 days no-questions-asked refund policy
✔ Language of communication will be strictly english
If you have questions or need any clarifications before enrolling, please
- refer to the FAQs at the bottom of this page
- if you still have questions, reach out to me at arpit.masterclass@gmail.com
Why should you join?
The primary objective of this program is to make you comfortable at building systems that are scalable, fault-tolerant, and reliable. But here is what you could reap out of it.
Build real AI systems end to end
Go beyond concepts and learn how to design, implement, and ship applied AI systems that work in real production environments.
Practical AI engineering
Learn the tooling, patterns, and engineering tradeoffs required to build robust AI applications.
Think in systems, not prompts
Develop the ability to design complete AI systems including data flow, evaluation, reliability, and failure handling.
Reusable patterns and architectures
Learn proven design patterns and architectures that you can directly apply to your own projects at work.
Hands-on, implementation first
Every concept is backed by code and prototypes so you can see how things work in practice.
Stay ahead in applied AI
Build the skills needed to work effectively with rapidly evolving AI tools and frameworks in industry.
Level up as an AI engineer
Position yourself to take on high-impact AI projects and stand out as someone who can deliver, not just experiment.
Community and peer learning
Join a group of serious engineers building AI systems and learn from shared experiences and discussions.
The world is learning
People from all over the world have taken my courses.
2000+
ENGINEERS
ENGINEERS
30
COHORTS
COHORTS
28
COUNTRIES
COUNTRIES
Who Took My Courses?
Folks belonging to some of the best companies and high thriving startups have taken this and other courses, the list includes the likes of
Teaching style
Here are some of the videos that will give you a peek into my teaching style how I teach and the depth I go into
Hey, I am Arpit
engineering, databases, and systems. always building.
I am a software engineer and engineering leader passionate about applied AI and databases. Currently, I am a Principal Engineer II at Razorpay, working at the intersection of Data and AI and building Agent Studio, which enables merchants to create AI agents while we provide the infrastructure and the harness to run them reliably.
Previously, I was a Staff Engineer at Google, where I worked on GCP Memorystore and GCP Dataproc. On the side, I am building - DiceDB, a fork of Valkey with multi-tiering and query subscriptions. I also spend time on independent research, publishing my work on ArXiv. My areas of interest include databases, approximate algorithms, and distributed systems.
In 2024, I took a leap of faith and co-founded Profile.fyi, which was later acquired by Mercor. I was part of Amazon's Fast Data Team, where I worked on cold tiering of hot data and developed a seamless query interface across all storage tiers.
I held engineering leadership positions (both IC and management) at Unacademy, where I built, grew, and led Search, Site Reliability Engineering (SRE) teams, and Data Engineering teams. I hold a total of 12+ years of experience in scaling backend services, taking products and teams from 0 to 1, and beyond.
I keep diving deep into engineering details and share my learnings through 166+ blogs, socials and, 250+ no-fluff engineering videos on YouTube, breaking down System Design, Database Internals, Microservices, BitTorrent Internals, Redis Internals, Research Papers, and the trade-offs behind real-world production systems.
Why Learn From Me?
- I lead Agent Studio at Razorpay, giving me hands-on experience building production AI systems
- I teach from first principles and focus on building strong engineering intuition
- I have already taught 2000+ engineers, senior engineers, and tech leaders
- My teaching style is highly engaging, practical and hands-on
- I focus on skills and systems that you can directly apply at work
- Everything I teach comes from real production experience, not theory or guesswork
What worked the best?
From the reviews and feedbacks I gathered, here are a few key things that worked for folks who took my courses.
mental models and frameworks
structured and well organised
quality and non-repetitive content
minute implementation-details
open ended discussion
much more than blogs we find
Frequently asked questions
You can always drop me an email at arpit.masterclass@gmail.com for other questions.
What will be the language of communication and teaching?
I will be teaching the entire course in english and all the brainstorming with participants will be conducted in english.
Is this course right for me?
This course is for any engineer who wants to learn the topics in the curriculum. Please go through the curriculum and decide if it covers what you want to learn.
Will you be giving teaching for the entire duration?
Yes. I will be teaching the entire course online and live over Zoom.
Will there be a class every day?
The Live Classes will happen on Saturdays and Sundays as per the time mentioned on the webite with a possible extension of 30/45 mins.
Can I get this course reimbursed from my company?
Talk to your manager and check if they can sponsor this course. The invoice that will be issued is a legally valid and sound invoice that can be used for any kind of reimbursement. Note: I will just share the invoice and certificate with you; and I will not be involved in any kind of reimbursement process, communication, or follow-ups from your finance and legal teams, it is between you and your employer.
When will I get the invoice?
You can download the invoice from the course portal. No invoice will be issued if you claim your refund. If you claim a refund, no invoice will be issued.
What if I want the invoice in the name of my employer?
If you want invoice in the name of your employer or a business for them to claim GST Input Credit (not TDS), drop me an email at arpit.masterclass@gmail.com. Note: I will not be involved in any kind of reimbursement process, communication, or follow-ups from your finance and legal teams, it is between you and your employer. I will not be dealing anything w.r.t TDS.
Will there be 1:1 mentorship sessions?
No. There will not be any 1:1 mentorship sessions for this course.
Will there be assignments and hands-on projects?
Yes. I would recommend you implement the core of every single system we discuss ensuring you apply what you learn.
Are assignments and projects mandatory?
No. But it is advisable that you complete them to get a better understanding of the system, algorithm, and business logic.
Will we also implement and see the systems in action?
Due to time constraints, it is not possible to implement every system; it is recommended that you self-implement the system and understand the low-level details. The course will definitely cover systems from every aspect.
Will there be a recording available for future reference?
Every single Live Class will be recorded, and you will be given lifetime access to it.
Is there a refund policy?
2 days no-questions-asked refund window from the course commencement date (11:59:59 pm IST on 1st Sunday after the course commencement. If you are moving from one cohort to another, you will become ineligible for the refund. To get a refund, you need to write to me at arpit.masterclass@gmail.com. If you somehow have the payment link of the live cohort and you enroll after the cohort has already started, you become ineligible for a refund.
Will I get access to other cohorts?
No. You will get lifetime access to the cohort you are part of and its recordings, or the recordings that you purchased.
Where are the classes conducted?
All Live Classes will be online, over Zoom, and all you need is an internet connection to attend the live sessions.
Will you be the only one teaching this course?
The entire course including Live Classes will be conducted by me, Arpit Bhayani, no external TAs, mentors, etc. You will get to learn everything from the horse's mouth.
What are the programming pre-requisites?
The course prerequisites are mentioned on this page.
Will I be getting an invoice of Payment?
Yes. An invoice will be issued to you with all the legal and necessary details. This means your employer can choose to process this invoice and provide reimbursement.
Will you issue a course completion certificate?
Yes. You can download it from the portal. The moment you download the invoice from the portal, you become ineligible to claim a refund, even if your refund window is not over.
Can I use my Credit Card or avail EMI to make the payment?
Yes, we support Credit Card, Debit Card, UPI, and Credit Card based EMIs having a duration of 3 months, 6 months, 12 months, and 24 months as offered by Razorpay.
Can I share the account with multiple people?
I track the browsers and devices from which the course is being accessed and if I fnd anything suspicious, I hold the complete right to revoke the access of the course and not offer any refund.