a hands-on course, one customer from the first meeting to production
Become a Forward Deployed Engineer: Ship AI for Real Customers
- Bramfield's systems on AWS
- messy data and file drops
- a claims triage agent
- an adjuster approves
- ₹4,444 or $149, lifetime access
- early access: 8 lessons ready, about 6 hours
A forward deployed engineer sits with one customer and gets AI working inside that customer's own systems. In this course you do that job once, from start to finish, for a fictional insurer whose systems you build on your own AWS account. Every step is the real work: the questions in the first meeting, the old systems, the messy data, the agent, the proof, the security review and the handover.
One payment, lifetime access. It rises to ₹9,999 or $249 once every part is released, and everything published later is included.
Early access: 8 lessons are ready to study today, about 6 hours of learning, with 4 hands-on labs, 3 of them in your own AWS account, and more than 150 diagrams and real screenshots. New lessons are added as they are finished, and you get every one. The price rises to ₹9,999 or $249 when the course is complete.
- the customer
- Bramfield Mutual, a fictional US insurer
- where the labs run
- your own AWS account, in us‑east‑1
- free to read first
- 4 lessons, with their labs
- Bramfield running
- about $0.03 an hour, priced for our real run
- the whole course on AWS
- about $8 to $12, estimated
- built from
- 41 job postings from 18 companies, read 13 November 2026
Why companies hire forward deployed engineers, and why the job is mostly delivery
Become a Forward Deployed Engineer: Ship AI for Real Customers is a self-paced, hands-on course by Roni Das on systemdesign.academy. It teaches the job of a forward deployed engineer by doing it: you work one customer engagement from the first meeting to a system the customer runs without you. The customer is Bramfield Mutual, a fictional US insurer with an old claims database, a SOAP policy service, nightly file drops and strict rules on personal data. You run discovery, measure a baseline, get into its systems safely, clean and mask its data, build a claims triage agent that an adjuster approves, prove it works with evaluations and shadow mode, put it into production in the customer's AWS account, pass the security review and hand it over. The labs run on real AWS in your own account. It is for engineers who know Python and SQL. No AI, AWS or consulting experience is needed.
Palantir describes the difference this way on its own blog (8 April 2019): "You can think of a Dev's focus as 'one capability, many customers,' while a Delta's focus is 'one customer, many capabilities.'" A Delta is what Palantir calls a forward deployed engineer. OpenAI's head of forward deployed engineering wrote on 22 October 2026 that he started the team in early 2025 "to close what we call 'the deployment gap'". In plain words, that is the distance between a model that works in a demo and a system that works inside a real company.
We read 41 forward deployed engineer job postings from 18 companies on 13 November 2026, from the companies' own public job boards. Production deployment and reliability came up in 37 of them. Scoping and discovery came up in 32, and so did AI agents. Fine-tuning a model came up in one. So this course spends most of its lessons where the job does. That means the customer's systems and data first, then proving the result, then running it in production. The model is one part of the work, not most of it.
The hard part is rarely the AI. It is the file that arrives at 2 a.m. with a renamed column, the old service that allows ten requests a second, the proxy that breaks every certificate, and the security team that asks where the data lives. None of the public pages we read for other forward deployed engineer courses, on 13 November 2026, mentions file drops, incremental loads, safe write-back, a corporate proxy, a security questionnaire or shadow mode. This course teaches each of them on real systems.
One claim through the pilot you build, from Bramfield's work queue to an adjuster's yes
This is the heart of the course. A claim arrives in Bramfield's work queue. Your code masks the personal data before any model sees it. The claims triage agent you build reads the claim, the policy form and the claim history through read-only tools, and drafts a decision with the reasons. An adjuster approves it, changes it or refuses it, and only then does your code write the result back to the queue, once, even if the call is retried. Every step is recorded, so months later anyone can see which file and which rows a decision used.
Bramfield's AWS account · us-east-1
- queueA claim arrivesa new auto claim lands in Bramfield's work queuethe claim and its notes
- PresidioPersonal data is maskednames, phone numbers and ID numbers, before any model sees themmasked text
- LangGraphThe triage agent readsthe claim and the policy form, through read-only toolsa draft decision
- adjusterA person decidesan adjuster approves, edits or refuses the draftapproved
- queueWritten back onceto the work queue, with no duplicate even when a call is retriedevery step
- The audit trailwhich file and which rows each decision used, kept where nobody can edit it
evalsBefore an adjuster sees any of it: the agent is scored on the golden set after every change, then runs silently beside the adjusters in shadow mode.
your path through the course
From the first meeting with a customer to a system its team runs without you
Each step below is a real part of the course, with how many lessons it holds. You work the same customer the whole way, so every document you write uses numbers you measured on its systems earlier in the course.
- Part 1The job and the customer6 lessons
- Part 2Discovery and scoping7 lessons
- SFTPPart 3Getting into the customer's systems9 lessons
- PIIPart 4From messy data to safe data6 lessons
- LangGraphMCPPart 5The pilot: an AI that does the job10 lessons
- evalsPart 6Proving it works6 lessons
- FargateauditPart 7Production inside the customer's walls9 lessons
- Handover, adoption and the next engagementHandover and the next engagement: 7 lessons, ending in a capstone you run on your own.
What you build in this course, for one customer, from the first meeting to the handover
The course is one engagement told in 8 parts. Each part leaves you with something you can show: a document with numbers you measured, or a running system on AWS. You start every lab with one command and stop or destroy it at the end, so nothing is left running.
- Bramfield's systems on your own AWS account, with its claims database, SOAP policy service, nightly file drop, documents, work queue and identity provider, built with one command.
- A measured baseline and a written scope of work, from Bramfield's own claims data: days to settle, the backlog, supervisor referrals and what each day of delay costs in dollars.
- Working connections into old systems, with paging and rate limits on the SOAP service, a file drop that survives messy rows, an incremental load without the hidden bug, and write-back that never makes duplicates.
- Clean, masked and traceable data, with personal data found and masked before any model sees it, and a golden set of claims whose answers the claims lead agreed.
- A claims triage agent, built with LangGraph, reading through read-only tools and an MCP server, pausing so an adjuster approves every decision, and holding against an instruction hidden in a claim note.
- Proof that it works, with an evaluation harness, an LLM judge checked against the claims lead, two simulated weeks of shadow mode, and a pilot readout with the failures stated plainly.
- Production in the customer's account, with private services on ECS Fargate, traces that never leave, a cost line for every claim, a rehearsed provider outage, a security questionnaire answered with evidence, and an audit trail.
- A handover and the next engagement, with the runbook, the field report back to product, a reusable pattern, an expansion proposal in dollars, and a capstone you run on your own.
What you leave the course with
- A portfolio of one complete customer engagement: the documents, the code and the measured results, in your own repository.
- The working habits the job postings ask for: scoping, integration, evaluation, production and handover, each practised on real systems.
- Practice for the forward deployed engineer interview rounds: decomposition, the customer conversation, debugging and design.
What the AWS labs cost you, and how the course keeps the bill small
Every lab runs in your own AWS account in us-east-1, and nothing runs on ours. The free lesson "Before you start" checks your account and sets a budget alert before anything is built. Apart from the hourly cost of Bramfield's systems, which we priced for our real run, the numbers below are the course plan's estimates from AWS list prices read on 13 November 2026; each lab lesson shows the cost we measured when we ran it, and these figures will be replaced by measured ones as the lessons are published.
- Bramfield's environment, which Parts 1 to 4 use, costs about $0.03 an hour while it is running, priced from the AWS Price List for exactly what our real run created. There is no NAT gateway on the main path, which keeps the hourly cost down.
- Over the whole course, about $8 to $12 of AWS in total (estimated), plus about $5 a month for stopped disks while you are working through it, if you stop the environment between sessions.
- Model calls are estimated at under $10 for the whole course, because most calls are short and the evaluations run a few hundred claims at a time.
- Every lab lesson ends with a stop or a destroy, a check lists anything still running, the Bramfield lab server, which Parts 1 to 4 use, stops itself four hours after every start, so a forgotten session costs little, and a $10 monthly budget emails you at $5 and $10 spent.
- Parts 1 to 4: Bramfield's systemsabout $0.03 an hourpriced for our real runOne small EC2 server (a t2.small) running the claims database, the policy service, the file drop, the work queue and Keycloak, with a public address and a 20 GB disk.
- Parts 5 and 6: the pilotabout $0.05 an hourestimatedThe same, plus a small RDS PostgreSQL database (db.t4g.micro) for the retrieval index, the audit table and the cost log.
- Part 7 and the capstone: productionabout $0.20 an hourestimatedThe same, plus two ECS Fargate tasks behind an internal load balancer, before its capacity units and log ingestion, which are small at course traffic.
The whole course on AWS: about $8 to $12, plus about $5 a month for stopped disks while you are working through it (estimated).
- 14 lessons reading and a knowledge check
- 4 lessons a real model API only
- 26 lessons real AWS in us-east-1
- 19 lessons real AWS and a model API
Estimates from AWS on-demand list prices for us-east-1, read from the AWS Price List API on 13 November 2026. Each lab lesson shows the cost we measured when we ran it. Lesson counts are from the course contents.
The contents of the Forward Deployed Engineer course, part by part, and which lessons are free
The course is planned as 63 lessons: the opening lessons and 8 parts. Pick a part, then open any lesson to see what it teaches and where its lab runs.
The 4 free lessons
- Introduction: what this course buildsfree, read now
- Before you startfree, read now
- What a forward deployed engineer doesfree, read now
- Meet Bramfield Mutual, the customer in this coursefree, read now
How the course is released. The course is released part by part, in book order. 8 lessons are ready to study today, about 6 hours of learning, with 4 hands-on labs, 3 of them in your own AWS account, and more than 150 diagrams and real screenshots. New lessons are added as they are finished, and you get every one. They cost nothing extra for anyone who has bought the course.
What the job really is, read from the companies that hire for it, how it differs from the roles around it, and the customer whose systems you inherit for the rest of the course.
One customer, many capabilities: how Palantir, OpenAI, Salesforce and Ramp describe the work in their own words, and what a real week looks like.
Where the lab runs: reading and a check
Open this lessonWho writes production code, who owns the outcome and who leaves after the slides, read from real job posts.
Where the lab runs: reading and a check
Open this lessonA fictional Pennsylvania insurer with a claims backlog, an old claims database, nightly file drops and strict rules on personal data; you build its systems on AWS and read the brief.
Where the lab runs: hands-on in your own AWS account
Open this lessonDiscovery, pilot, production, handover and expansion, what each phase must produce, and the pod of two engineers and one strategist.
Where the lab runs: reading and a check
Open this lessonMapping Bramfield's servers, databases, file drops and services with read-only access, and drawing the system map you will use all course.
Where the lab runs: hands-on in your own AWS account
Open this lessonThe three documents that let anyone else pick up your account, kept in Git from the first day.
Where the lab runs: reading and a check
Open this lesson
Who this forward deployed engineer course is for, and what you should know before you start
- Backend, full-stack and data engineers with one to eight years of experience who want a forward deployed engineer job and want to practise the work before the interview
- Solutions engineers, implementation engineers and consultants who want to own the production system, not only the demo
- Engineers already in customer-facing AI roles who want a structured method and one complete engagement to show
Before you start, you should know
- Python: you can read and change a 200-line program
- SQL: you can write a query with a join
- The terminal and Git
- An AWS account you can put a card on
You do not need to know
- AI experience: each idea is explained the first time you need it
- AWS experience: the course builds everything with one command and explains what it built
- Consulting experience: the customer lessons show you what to ask and what to write
How the course is delivered: written lessons, real runs on AWS, and a check after each idea
Every lesson written out in full
Each lesson starts with a real problem at Bramfield, explains why it matters, draws the idea, and then does it on real systems.
Real runs on real AWS
What you see in a lesson comes from a real run in us-east-1, with its date and its cost, and the lesson ends by stopping or destroying what it started.
Where each lab runs, stated up front
Every lesson in the contents says whether it is reading only, calls a model API, runs on AWS, or both, so you know before you start.
A fictional customer, synthetic data
Bramfield Mutual does not exist. Every record in its systems is made by the course's own generator, and no real person's data is used anywhere.
The early-access price of Forward Deployed Engineer, and why it is lower now
The course is released part by part, so it sells at an early-access price today. Once every part is released, the price rises to ₹9,999 or $249. If you buy now, you pay once and get every part as it comes out, at no extra cost.
Early access: 8 lessons are ready to study today, about 6 hours of learning, with 4 hands-on labs, 3 of them in your own AWS account, and more than 150 diagrams and real screenshots. New lessons are added as they are finished, and you get every one. The price rises to ₹9,999 or $249 when the course is complete.
Not sure yet? Study the free lessons first.
See every lesson in the contentssystemdesign.academy
receipt · Forward Deployed Engineer
- 63 lessons in 8 parts
- included
- Parts published later
- included
- Code zips
- included
- Full price, later
- ₹9,999 or $249
- You save today
- ₹5,555 or $100
Total, one payment₹4,444 or $149
lifetime access
Buy early accessIn India you pay in rupees through Razorpay, by UPI, card or netbanking. Outside India you pay in US dollars by card through Razorpay, or with PayPal.
Questions people ask before buying the Forward Deployed Engineer course
Who is this forward deployed engineer course for?
It is for backend, full-stack and data engineers who want a forward deployed engineer job, for solutions engineers, implementation engineers and consultants who want to own the production system and not only the demo, and for engineers already in customer-facing AI roles who want a method and a portfolio. It is written for readers anywhere in the world, and every money figure in it is in US dollars.
Are all the parts of Forward Deployed Engineer available on the day I buy?
No. The course is planned as 63 lessons. The course is released part by part, in book order. 8 lessons are ready to study today, about 6 hours of learning, with 4 hands-on labs, 3 of them in your own AWS account, and more than 150 diagrams and real screenshots. New lessons are added as they are finished, and you get every one. They cost nothing extra for anyone who has bought the course. That is why it sells at an early-access price, ₹4,444 in India or $149 elsewhere, which rises to ₹9,999 or $249 once every part is released. Everything published later is included in the one payment.
What do I need to know before I start?
You should be able to read and change a 200-line Python program, write a SQL query with a join, and use the terminal and Git. You also need an AWS account you can put a card on. You do not need experience with AI or AWS, and you do not need to have worked as a consultant. Each idea is explained the first time you need it.
What will the AWS labs cost me?
The course plan estimates about $8 to $12 of AWS for the whole course, plus about $5 a month for stopped disks while you are working through it, if you stop the environment between sessions. Model calls are estimated at under $10 for the whole course. These are estimates from AWS list prices for us-east-1 read on 13 November 2026, not measurements yet. Each lab lesson shows the cost we priced for what our run created, and these numbers are replaced by the measured ones as the lessons are published. The free lesson "Before you start" sets a $10 budget alert before anything is built, the Bramfield lab server, which Parts 1 to 4 use, stops itself four hours after every start, and every lab lesson ends by stopping or destroying what it started.
Is Bramfield Mutual a real company?
No. Bramfield Mutual Insurance Company is a fictional US insurer made up for this course, and any likeness to a real company is accidental. Every policy, claim and document in its systems is synthetic, made by the course's own generator, and no real person's data is used anywhere. Even the identity numbers are made so that they can never be real: every Social Security style number uses a group number that the Social Security Administration says it will never assign.
What do I get if I buy early access?
You pay once and get lifetime access to every lesson of the course as it is published, with the code zip for each lesson that has code, at no extra cost. You can study the free lessons before you buy, and the contents page shows which lessons are ready today. The early-access price is lower because the course is still being released, and it rises once every part is out.
How is this course different from a live forward deployed engineer bootcamp?
This course is self-paced: there are no live sessions to attend and no start date to wait for. It follows one customer engagement from the first meeting to production, and the main path runs on real AWS in your own account, with a teardown at the end of every lab lesson. It also spends whole parts on the integration work that takes most of an engineer's first weeks with a customer: file drops, incremental loads, safe write-back, a corporate proxy, a security questionnaire, shadow mode and tearing the cloud resources down. If you learn best with a live teacher and a group around you, a live programme may suit you better.
Is this the same as the free forward deployed engineer career guide?
No. The free career guide at /forward-deployed-engineer explains the job: what it is, what the companies ask for and how people get there. This course is where you practise the job, on one customer, from the first meeting to production.
Is Forward Deployed Engineer a video course or a text course?
It is a written course, laid out like a book of parts and lessons. Every lesson is written out in full, with real screenshots from real runs, a diagram for each idea and a short knowledge check, and each lesson with code comes with a zip of that code. If videos are added to the lessons later, they are included at no extra cost.
Is Forward Deployed Engineer included in the System Design, AI Engineering or AIOps courses?
No. It is its own course with its own price. Buying it gives lifetime access to all of its lessons and code, including every part published later.
How do I pay for Forward Deployed Engineer?
In India you pay in rupees through Razorpay, by UPI, card or netbanking. Everywhere else you pay in US dollars by card through Razorpay, or with PayPal. The course unlocks on your account as soon as the payment is confirmed.
