Forward deployed engineer roadmap: 7 phases, 7 proof projects

Most people who ask me for an FDE roadmap are not short of resources. Their worry is different. They fear spending six months on videos and still not knowing whether they are ready, or starting from the wrong place and wasting the first three. I teach the FDE fresher track at NeuraPath, so I see where people actually stall, and it is rarely the topic they were nervous about. This is the roadmap I would give a friend. It has two starting points, a time budget, and a gate at the end of every phase: a project you must ship before you move on. If you want to know which starting point is yours before reading further, the free 5-minute resume check will tell you.

How I put this together: the phases, projects and week ranges come from our own fresher syllabus. The market figures come from the 138 FDE and related roles in our India jobs database on 12 Oct 2026. That database is the set we track from employers' careers pages, not every FDE job in India.

Where should you start, as a fresher or as a working engineer?

Start at Phase 1 if you have never shipped code that someone else installed and ran. Start at the Phase 3 gate if you write and deploy code for a living, and treat the first two gates as a test you take, not a course you sit through.

The honest test for a working engineer is simple. Try to build the Phase 1 and Phase 2 proof projects from the table below. If each takes you a weekend or two and you would be comfortable showing the code to a stranger, skip ahead. If either takes a month, or you find yourself copying a tutorial line by line, do the phase properly. Years on a CV do not change this. The gate measures what you can ship today, not how long you have been employed.

Freshers should also know what the market looks like at the far end. Of the 43 postings in our database that state a minimum experience, only 2 ask for a year or less, and the median is 5 years. That does not mean a fresher cannot get there. It means the roadmap has to end with proof strong enough to stand in for experience, and that your first role may be an adjacent title on the way. I cover that in FDE jobs for freshers in India.

What are the phases, and what project proves each one?

Seven phases, in this order, each closed by a project someone else can run. The week ranges are from our fresher syllabus, which spreads about 280 learning hours (live sessions plus self-study and project time) across roughly 30 weeks.

Two things about this order are deliberate. Deployment comes before AI because an agent you cannot deploy, secure or monitor is a demo, and clients do not pay for demos. And each project builds on the last: the Phase 3 project is the Phase 2 service, made production-grade. By Phase 7 you have one connected body of work, which reads far better in an interview than seven unrelated tutorials.

Why does every phase need a proof project?

Because understanding a topic and being able to ship it are different skills, and the gap only shows when someone else runs your code. A gate turns a vague sense of progress into a yes or no.

I saw this in our own class. In the 12 Sep live session, the notebook we typed in the room hand-wrote a logfile.txt that had three bugs in five lines, and it hardcoded a C:\Users\… Windows path that would break on anyone else's machine. It ran. On that laptop, on that day. We built the next two sessions, on 13 and 19 Sep, around fixing exactly that: turning the notebook into a package with modules, __main__, __init__.py, a virtual environment and the logging module, then making it trustworthy with docstrings, type hints and pytest. Code that works on my laptop is not FDE code, because an FDE's code always ends up on someone else's machine.

So a phase is done when its project passes all of these, not when you finish the videos:

• A stranger can clone it, follow the README and run it without messaging you.
• There are tests, and they pass on a clean machine.
• No path, password or key is hardcoded.
• Someone other than you has reviewed the code, even a friend from college.
• You can explain, in two minutes, one decision you made and what you would change.

How many hours a week does it take, and for how long?

At 9–12 hours a week, a fresher following this order should plan on about 30 weeks, which is six to seven months. A working engineer who tests out of phases 1–3 can do it in far less, and our proposed FDE for Professionals track is planned at 16 weeks for that reason.

The time goes into the gates, not the videos. A useful split for a week of about ten hours: three hours learning something new, five hours building the phase project, and two hours on review, either reading someone else's code or fixing comments on yours. Our weekend cohort runs live from 10:00 AM IST on Saturday and Sunday for three hours each, and the rest of the week is project work, which is roughly that shape.

Below about six hours a week, I would not try this. You forget Phase 1 before you reach Phase 3, and the project chain breaks. If your job or college genuinely leaves you less than that, it is better to wait a few months and start properly than to start thin. Self-study also tends to run slower than the table, for one reason: nobody tells you when a gate has failed, so people carry weak foundations forward and pay for it in Phase 4.

Should you learn Python or SQL first?

Python first, then SQL from around week 8. Phase 1 is where you learn to structure a program, test it and use Git, and every later phase depends on that. SQL is easier to add once you can write the service that runs the query.

Look at what postings name. Of the 138 roles in our database, 36 mention Python and 35 mention APIs, while 15 mention SQL. Do not read that as "SQL does not matter". Almost every integration an FDE builds touches a database, and postings often leave SQL unsaid because they assume it. The order matters more than the count: Python, then SQL inside a real service, then the API around it. If you already write SQL daily as an analyst, you still start at Python. Your SQL will make Phase 2 quick.

Do you need DSA for an FDE role?

You need the core, not a competitive-programming grind. Arrays, strings, hashmaps, stacks and queues, recursion, Big-O intuition, and two common patterns (two pointers and sliding window) are enough for most FDE coding screens.

That is what our Phase 1 covers in a 13-hour problem-solving module, with short drills that continue weekly through the rest of the programme. I keep it small for a reason. FDE postings talk about client work and AI far more than algorithms: 67 of our 138 listings mention client-facing work and 55 mention AI agents. You will still meet a coding round, and failing it ends the loop, so do the drills. But four hundred LeetCode problems will not help you in the round where a client asks why your assistant gave a wrong answer. For what each round actually tests, see FDE interview questions with model answers.

Can you follow this roadmap for free?

Yes. Every technical phase has good free primary material, and I would rather you start free than not start. What free cannot give you is feedback on your work and practice with a real person playing the client.

• Phase 1: the official Python tutorial, the free Pro Git book, and the pytest documentation. Publish the CLI tool and ask one person to review your pull request.
• Phase 2: SQLite or PostgreSQL with their own docs, the FastAPI documentation, and any public API you find interesting.
• Phase 3: Docker's own getting-started guide, GitHub Actions on your repository, and a cloud provider's free tier for the live URL.
• Phase 4: the LLM providers' official documentation and cookbooks, and the RAGAS docs for evaluation. Use a document set you know well, such as your college's rulebooks.
• Phase 5: the MCP specification at modelcontextprotocol.io and the LangGraph documentation.
• Phase 6: this is the hard one to do free. Ask a friend to play a difficult client, record the call, and score yourself against the transcript in our customer-scenario interview round.
• Phase 7: pick a fictional company, write its discovery document, build and deploy, then demo it to someone non-technical and write the handover.

I wrote about where free material runs out in more detail in is there a free FDE course?. The short version: if your barrier is knowing what to learn, this page is enough. If your barrier is finishing, and getting honest comments on what you build, that is where paid teaching earns its money.

What does the taught version of this roadmap look like?

Our FDE for Freshers programme follows exactly this order over 6–7 months, and it needs no coding experience to start. It is taught live online by working practitioners, with optional classroom labs at our Noida centre, in batches of 15–20 learners, and sessions are recorded. I lead it. My day job is Senior Data Scientist at Parexel, a Goldman Sachs–backed company, after about ten years in data science and AI, and I have taught as a freelance trainer at BIA, AnalytixLabs and Scaler. Ishaan Sharma, a Senior Data Scientist at Nuvo AI (Meril Group) with about a decade in NLP, computer vision and GenAI, is our co-instructor for AI and GenAI.

Each phase project is code-reviewed one-to-one by a mentor. The Phase 5 project, for example, is an MCP server that exposes internal tools, plus a LangGraph agent that uses it, shipped with guardrails, an evaluation suite and observability. The fee is inclusive of GST, with EMI from ₹8,000/month; the full fee is shared on a free counselling call. Eligible learners are covered by our job guarantee with minimum package assurance, or 50% of the fee back, for those meeting the attendance, deliverable and capstone criteria. Terms and conditions apply. The conditions are plain: at least 70% in every assessment, at least 80% live attendance, at least four logged job applications a week, fees settled, and attending the interviews we arrange, within six months of certification. The minimum package and full conditions are in our published terms.

Who this is not for: if you already ship production code and only want the AI layer, the fresher track will bore you for its first four months, so look at the proposed professionals track or self-study phases 4–6. If you cannot give nine hours a week for seven months, wait until you can. And if you want a certificate more than a portfolio, any shorter course will suit you better.

What to do this week

1. 1Decide your starting point with the test above: try the Phase 1 project, and time how long it takes.
2. 2Create the GitHub repository for it today, and write the README before the code: what the tool does, how to install it, how to run it.
3. 3Block ten hours in your calendar for the next four weeks, and protect them like a class.
4. 4Read five real postings in the India FDE jobs report and note which skills from the table each one names.
5. 5At the end of the week, ask a friend to clone your repository on their laptop and run it. Fix whatever breaks.