Stop Building Like a Developer.
Start Thinking Like an Architect.

Yes, in Hinglish: technical terms in English, explanations in Hindi the way Indian engineering teams actually talk. It is a live system design course for backend engineers in India, not dubbed English slides. If you searched for a system design course in Hindi or Hinglish, this is built for that.

Yes. The whole arc is developer to architect: one teaching system, earn every split with a reproducible failure, write ADRs for the decisions that matter, and leave with a way of thinking hiring managers can hear. It is not syntax training. We build only what is needed to learn architecture and system design.

People who already build software, not beginners. You should have developed APIs, worked with databases, and shipped apps or services. This cohort is architectural thinking and system design in a structured way. It is not how to write a CRUD API. Sweet spot is mid-to-senior backend engineers who have felt a bad architecture decision in production.

You should already have developed APIs, worked with databases, and built apps or services. Comfortable cloning a repo, running Docker, and reading a REST contract. This is not for beginners. We do not teach syntax or how to stand up your first service. We teach architectural thinking and system design, in a structured way. You do not need an AI background. The AI weeks start from the basic words, and I send a short pre-read first.

All sessions are recorded and available to every enrolled student for life (watermarked, no public distribution). The recordings are unedited, so you see the full class, including the Q&A and the mistakes. There is no live coding. The live Q&A and the running demo are where the real learning happens. Try not to miss them.

Yes. There is no grade and we do not fail you for C#. Clone the branch, run the demo, follow the decisions. k6 shows up in Week 6, when we load-test. Tadka runs on .NET 10, but service boundaries, API contracts, Kafka, caching, and observability are the same in Java, Node, or Go. Every day ships an option-space.md that maps the pattern to your stack. The AI weeks work the same way: the ideas (untrusted model output, tool permissions, evals) are identical in any stack, and the code in the repo is C#.

Most system design and architecture courses give you twenty shallow interview diagrams. We take one teaching system and run the architecture arc: monolith, scale until something hurts, split when the data says so, load test it, break it on purpose. We build only what is needed to learn those moves, not a production food-delivery product. That helps in interviews because you have seen a cache stampede, not just drawn a Kafka box.

No live coding. No grades, no rubric, no homework marks. Tadka is C# in the repo so you can clone and run it. Java, Node, and Go people watch the same demo and write the same kind of decisions. Your job in class is the thinking, not typing handlers.

Everyone does pair mocks in Week 8, plus the Design-Swiggy masterclass. I cannot do forty-five minute 1:1s with fifty people and still teach the cohort. The top five peer-voted ADRs each batch get an instructor review slot.

No. We build only what is needed to learn architecture and system design. Postgres, Redis, Kafka, Docker, a bit of observability, k6, and ADRs show up when the teaching system needs them. The week list is the live spine. Break-kits in the repo are optional. MongoDB, Elasticsearch, event sourcing, and fancier deploy patterns come up as trade-offs, not as a second product. AI engineering follows the same rule: one small assistant inside Tadka, not a tour of every AI tool. No model training and no fine-tuning.

Very little. Postgres, Redis, Kafka, and Docker all run free on your laptop. Cloud deploy is one black-box day: we use a pre-built stack and talk about blast radius, release cadence, and the monthly bill. You do not iterate on AWS infra yourself, so no surprise cloud charges. For the AI weeks you do not need to buy anything. The required path is built to run with a free-tier model or a local model, and I demo on Azure AI. Free-tier limits change, so I confirm the current options before the batch starts.

About 5 hours of live sessions on weekends: 3 on Saturday and 2 on Sunday. Some classes have a short pre-read of 20 to 30 minutes, so live time goes to decisions and not to basics. Clone and run the branch if you want the demo on your laptop. ADRs in class are practice, not a graded assignment. Optional extra time is yours. It is 8 weeks, not a second job.

Weeks 6 to 8 add AI engineering to Tadka. Foundations and cost per request. Search with Postgres full-text and pgvector, and a semantic cache that never skips auth, price, or freshness. Tool calling and MCP, prompt injection, evals in CI, traces without personal data, prompt versions, model fallback, and a token budget. All of it on one small assistant. The rule through the whole block: the model suggests, your server decides. I skip model training, fine-tuning, and agent swarms.

You walk out with a way of thinking, the repo, and templates for ADRs, diagrams, and a cost readout. That is what hiring managers hear in a design round, not a PDF that says you completed a course. We do not grade a portfolio for you.

Attend the first live class (Day 1). If the teaching style or pacing is not right for you, email hello@desiarchitect.com before the second day starts and you receive a 100% full refund, no questions asked. Once Day 2 begins, seats are locked and no refunds will be provided.