In-person · 20 weeks · Bengaluru · Weekends · Starts November

Build the systems
that keep LLMs fast,
cheap, and ready to scale.

An in-person cohort on Inference Engineering & AI Infrastructure, led by engineers who run inference in production.

20Weeks

Across 20 weekends

In-person

Bengaluru, India

5Instructors

Active industry engineers

₹999

To apply — fully refundable

Why This, Why Now

Inference engineering is the fastest-growing discipline in AI, and the scarcest.

75%

Of all AI compute will go to inference by 2030, up from roughly half in 2025.

McKinsey, 2026

50–65%

Gross margins on AI-native products, versus 80–90% for SaaS. The gap is inference efficiency.

AI Startup Benchmarks, 2026

₹70L–1.6Cr

Total compensation for senior inference and GPU-systems engineers in India.

India Market Data, 2026

“The engineering discipline behind it is vastly underappreciated, poorly understood, and frankly, undersupplied with good engineers.”
Chris Zeoli · The Rise of Inference Engineering

The Cohort · Instructors

Engineers who've built AI infrastructure & inference systems at scale.

Tarun Tiwari

GPU Cloud & Infrastructure

Tarun Tiwari

Head of Product · Nava

Built the Alexa AI platform behind 100M+ devices. Now builds the GPU cloud and training-cluster infrastructure that AI teams run large models on.

MIT · IIT Kharagpur

Ashish Prasad

LLM Systems & Retrieval

Ashish Prasad

Eng. Manager, AI/ML · Microsoft Turing

Helped put GPT-5 inside Microsoft 365 Copilot, working with a team of 30 across engineering and research.

NIT Karnataka

Shubhendu Shishir

Inference & Serving

Shubhendu Shishir

Head of Engineering · Simplismart

Runs the engineering behind one of the fastest inference engines in production.

NIT Jamshedpur

Naman Vats

Agents & Evaluation

Naman Vats

Senior Software Engineer · Sentient Labs

Published research on why the same AI model can cost 40× more to run depending on how you build around it.

BIT Mesra

Mayank Sharma

Model Training & Fine-Tuning

Mayank Sharma

AI Lead · Opkey

Trains models from scratch, and has led ML teams building for images, language and on-device AI.

UT Austin · BITS Pilani

Phase-by-Phase Curriculum

What you'll learn.

Five phases across 20 weekends, taking you from how a model executes to a production serving system you defend in front of a panel.

Phase A · 01 / 05

Model and Measurement Foundations

Weeks 1–4

Understand how a model executes, then establish the quality and performance baselines every later optimisation is measured against.

  1. 01LLM execution under the hoodUnderstand how an LLM actually works inside.Week 1
  2. 02Autoregressive inference mechanicsSee where every millisecond of a response goes.Week 2
  3. 03Evaluation and quality baselinesDecide what "good output" means, and measure it.Week 3
  4. 04Performance benchmarking and profilingPut hard numbers on speed, cost and throughput.Week 4

Phase B · 02 / 05

Hardware Prerequisite

Weeks 5–6

Reason from GPU compute, memory and interconnect constraints before touching engine internals.

  1. 05GPU compute architecture for inferenceLearn what the GPU is really doing.Week 5
  2. 06GPU memory, interconnects and topologyKnow where memory goes and when it runs out.Week 6

Phase C · 03 / 05 · core

Inference Engines and Optimisation

Weeks 7–14

One continuous block from engine internals to measured single-node optimisation. The core of the program.

  1. 07vLLM deep dive I: serving and schedulingSwap the naive server for a real one.Week 7
  2. 08vLLM deep dive II: PagedAttention and KV memoryMake the KV cache stop wasting memory.Week 8
  3. 09SGLang deep dive I: runtime and RadixAttentionReuse what the model has already computed.Week 9
  4. 10SGLang deep dive II: workload-aware optimisationTune the engine to the traffic you actually get.Week 10
  5. 11TensorRT-LLM, deployment artifacts and TritonShip a compiled, optimised artifact.Week 11
  6. 12Quantisation and precision engineeringShrink the model without breaking it.Week 12
  7. 13Speculative decoding in depthGenerate several tokens for the price of one.Week 13
  8. 14Prefix, index and tiered KV-cache systemsCache across requests, GPUs and storage.Week 14

Phase D · 04 / 05

Distributed Inference and Platform

Weeks 15–18

Scale the runtime past one GPU, then build the serving platform that deploys, routes, observes and recovers it.

  1. 15Multi-GPU, MoE and distributed inferenceScale past a single GPU.Week 15
  2. 16Prefill-decode disaggregation and routingSplit prefill from decode, and route by cache.Week 16
  3. 17Platform engineering and model lifecycleGet it from a registry to a real endpoint.Week 17
  4. 18LLMOps, traffic management and reliabilityKeep it up when traffic spikes and things fail.Week 18

Phase E · 05 / 05

Bounded Extensions and Capstone

Weeks 19–20

Add the adaptation and application layers the job needs, then defend the whole system in front of a panel.

  1. 19Bounded adaptation and application workloadsFine-tune, serve adapters, and wire up retrieval.Week 19
  2. 20Capstone validation and technical defenceProve the whole thing works, and defend it.Week 20

Eligibility

Can you join?
Here's what we look for.

Your job title does not matter. If most of these sound like you, apply.

01

2+ years, full-time

Relevant professional work. Internships do not count.

02

ML or applied ML background

Or backend/platform, if you meet the same ML bar.

03

Professional Python

Modular code, testing, debugging, Git.

04

Hands-on PyTorch

Run and modify a model, inspect tensors, use autograd.

05

Transformers at working level

Attention, embeddings, tokenisation, how generation works.

06

Working systems knowledge

Linux, HTTP, concurrency, Docker.

Admission Process

Six steps to your seat.

We review applications on a rolling basis. Apply early — seats fill before the deadline.

  1. Pay ₹999 to apply

    Secures your interview slot. Fully refundable if you are not shortlisted for the interview.

  2. Fill the application form

    Tell us about your background, current role, years of experience, and why you want to join.

  3. Get invited for a 1-on-1 interview

    A 30-minute technical and fit interview with one of the instructors, typically within 5 days.

  4. Get selected

    We notify you within 24 hours of the interview. If selected, you receive a formal offer letter.

  5. Secure your seat

    Pay the ₹5,000 registration fee to confirm your place. Seats are confirmed in order of acceptance.

  6. Complete the admission process

    Review the program agreement, set up your tools, and join the cohort Slack before Day 1.

Program fee

Program fee and
what it covers.

Pay ₹999 to apply — the rest is due only once you have an offer and have accepted your seat.

Application fee

₹999

100% refundable if not shortlisted for the interview

What the fee covers

  • 240 guided hours across 20 weekends
  • Live weekend sessions with practitioner instructors
  • GPU compute for every lab, benchmark and capstone run
  • Capstone mentorship through to submission
  • Session recordings — lifetime access
  • Project and code reviews throughout

Program fee is refundable as per our Refund & Withdrawal Policy. Seats are confirmed in order of acceptance.

You pay

₹2,06,500

Inclusive of GST · Single payment

Registration fee(non-refundable)

₹5,000

Program fee(refundable per policy)

₹1,70,000

GST (18%)

₹31,500

You pay

Instalment 1

₹50,000

Instalment 2

₹1,56,500

Flexible plans, no financial stress.

Registration fee(non-refundable)

₹5,000

Program fee(refundable per policy)

₹1,70,000

GST (18%)

₹31,500

You pay

₹37,613/ month × 6

Spread the cost over 6 months.

Registration fee(non-refundable)

₹5,000

Program fee(refundable per policy)

₹1,70,000

GST (18%)

₹31,500

Offline Experience

Take your learning beyond the screen

Step away from zoom screens. Build production-grade inference pipelines shoulder-to-shoulder with elite systems engineers.

Masai campus entrance
Students in a classroom session
Learners at work in the lab
Peer programming in progress
Mentor whiteboarding with a learner
Masai HQHSR Layout, Bengaluru
View on Map

Apply

Ready to run LLMsin production?

Applications for the November cohort are open.

Apply for the November Cohort

Selective admission · ₹999 to apply · Fully refundable if not shortlisted

FAQ

Frequently asked questions.

What does the fee include?

All 240 guided hours, live weekend sessions, project reviews, capstone mentorship — and the GPU compute you use throughout. Every lab, benchmark and capstone run happens on machines we provide, so there is no separate cloud bill. You just need your own laptop.

When do I pay?

In three stages. ₹999 when you apply, which secures your interview slot. If you are selected, a registration fee of ₹5,000 confirms your seat. The balance is due before the first weekend, and you can pay it upfront, in two instalments, or as 6 monthly EMIs through our NBFC partners.

Do I need a machine learning background?

Yes. This one is ML-first: you should already write production Python, work comfortably in PyTorch, and understand how transformers generate text. Backend and platform engineers are welcome, but they meet the same ML bar rather than substituting systems experience for it.

How much time does it take?

Six hours on Saturday, six on Sunday, plus about 30 minutes a day during the week. About 320–360 hours in total.

Can I keep my full-time job?

Yes. That is what the weekend format is for — we expect everyone to be working, and the weekday work is built to fit around a job.

What if I miss a weekend?

Sessions are recorded, so missing one is fine. Miss several in a row and your project deadlines slip, so we would work out a plan with you.

What do I finish with?

You will understand an open LLM, make it faster and cheaper, and run it as a reliable production service.

Is there a placement guarantee?

No. This is for engineers who already have jobs. What you get is proof you can run LLM systems in production — the four projects.