--- title: "What is a forward deployed engineer?" description: "What a forward deployed engineer does, how the role differs from a consultant or an agency, and when a fractional hire fits." canonical_url: "https://fractionalappliedai.com/forward-deployed-engineer" meta: "og:description": "What a forward deployed engineer does, how the role differs from a consultant or an agency, and when a fractional hire fits." "og:image": "https://fractionalappliedai.com/_og/s/o_ecd0229a12e8c58a.png" "og:image:alt": "What is a forward deployed engineer? | Fractional Applied AI" "og:image:height": "630" "og:image:type": image/png "og:image:width": "1200" "og:title": "What is a forward deployed engineer? | Fractional Applied AI" "twitter:image": "https://fractionalappliedai.com/_og/s/o_ecd0229a12e8c58a.png" "twitter:image:alt": "What is a forward deployed engineer? | Fractional Applied AI" --- > Index of every page on this site: [llms.txt](https://fractionalappliedai.com/llms.txt) # What is a forward deployed engineer? A forward deployed engineer is a senior engineer who works inside your company, on your systems, until an AI feature is running in the workflow it was meant for. The deliverable is software your team uses after the engagement ends. Companies search for the title because they are trying to name a gap. A consultant can recommend an approach. An agency can deliver a project from the outside. A full-time search can take months. A forward deployed engineer sits between those options: one experienced person, embedded for a bounded stretch of time, responsible for shipping. This page explains the role, how it differs from a consultant or an agency, when it is the right hire, and how the cost of fractional work compares with a full-time engineer. The [services](https://fractionalappliedai.com/services) page lists what I ship. The [contact](https://fractionalappliedai.com/contact) page is where you request an intro call. ## What the person actually does The engineer sits with the people who do the work. That might be an operations lead, a support queue, a finance clerk, or the team that already owns the product. Together you name one job: a decision that is slow, a handoff that drops information, or a product surface where a language model would help if it were wired to real data. Then the engineer builds that job into the system you already have. An agent that can read a record, take a bounded action, and leave an audit trail. A language model feature inside the product your customers already open. A workflow that used to be a pile of tickets and is now a path, with a human check at the step that needs one. The engineer uses your repository, your issue tracker, and your staging environment. The code stays in your account. When the engagement ends, your team can change it without calling anyone. I do this work as a fractional engagement. The same pattern exists as a full-time role at larger companies. The fractional version is for a company that needs the person for a few months, and does not want to open a permanent seat. ## Where the title comes from "Forward deployed" is older than the current wave of AI products. It described engineers who were sent to the customer, close to the operation, instead of building only from a headquarters roadmap. Palantir popularized the title for engineers who worked alongside the teams using the software. The name stuck because it described a stance: go to the work, learn the constraints, and build against those constraints. The title has since spread. Companies use it for engineers who implement AI in a customer's environment, and for senior people who will stay with a problem until it is in production. Some job posts use "forward deployed engineer" for a pre-sales role that configures a vendor product and never touches the customer's codebase. That is a different job. Read the description, not only the title. When I use the title, I mean the narrower thing. A forward deployed engineer writes and ships software in your stack, against your data, with your team, for a defined period. Applied AI is the subject matter. The stance is the same if the work is an agent, a model call inside an existing product, or an automation that removes a manual loop. ## Applied AI, named plainly Applied AI means the model is in a workflow, with inputs, limits, and a person who owns the outcome. An applied AI engineer is the person who can design that workflow, implement it, and sit with the failure cases until the behavior is acceptable on your data. A fractional AI engineer is that person, hired for part of a week or for a few months rather than as a permanent employee. Fractional applied AI is the practice: applied work, on a fractional engagement. The search terms differ. The work does not. You are hiring someone to put a model to work inside a real process, then leave the system in a state your team can run. The audience I write this for is a founder, a CTO, or an operations leader at a small or mid-size company. You have a real product or a real operation. You have felt the gap between a demo and something your staff will trust on a Tuesday afternoon. You want an embedded engineer for a period of time, and a full-time hire is the wrong shape or the wrong time. ## A normal week Early in an engagement, a large share of the time goes to reading. How a record is created. Who is allowed to change it. What "done" means for the person at the end of the process. Which numbers are trustworthy and which are folklore people repeat because the spreadsheet has always said so. After that, the week looks like any senior engineer's week. A branch. A small release. A review with the person who will live with the feature. A fix when the model is wrong in a way the first prompt did not predict. Time with the operator who hit the bad case, because that person has the example the eval set was missing. I keep the surface area small on purpose. One workflow, or one feature, until it is in production and someone other than me has used it on a real case. Then the next one, if the engagement continues. A wide program of five workstreams is a team. I will say so rather than pretend one person can staff it. The tools are yours. I work in the issue tracker you already have, I open pull requests your engineers can read, and I deploy to the environment your team already trusts. Status is the diff and a short note about what a real user did with it. A weekly meeting exists to look at that, and to decide the next slice. ## How this differs from an AI consultant A consultant's product is advice. A good consultant will interview your team, look at your systems, and write a recommendation: what to buy, what to build, what to ignore, and in what order. That work is valuable when you do not yet know which problem is worth software. A consultant is usually paid for meetings and documents. When the engagement ends, you hold a plan. Someone else still has to build it, and someone else will be holding the pager when the integration fails on a Friday. A forward deployed engineer starts from a similar conversation, because building the wrong thing quickly is still waste. The difference is the output. The output is a change in your product or your operation. If you need a recommendation, it is a short note attached to a working system. Hire a consultant when the decision is whether to do the work at all, or which vendor to trust, and you have people who will implement the answer. Hire a forward deployed engineer when you already know the job you want done and you need someone who will do it inside your stack. If you need the decision first, the advisory hours on the [pricing](https://fractionalappliedai.com/pricing) page are the engagement for that. The retainer and the sprint are the engagements for the build. ## How this differs from an agency An agency brings a team, a producer, and a delivery process. That shape fits when the work is larger than one person, when you want several specialties at once, or when you are standing up a new product surface from scratch. The trade is distance. The agency's team is working your project among other projects. Knowledge of your operation sits in tickets and status calls. Handover is an event at the end. The codebase they leave is often a separate application, because staffing a separate application is easier than learning yours. A forward deployed engineer is one person who joins the tools you already use. There is no account layer between you and the person writing the code. The work lands in your repository, in your design system, behind your auth. Reviews happen with your engineers, if you have them, or with you if you do not. The limit is the same fact. One person cannot staff a multi-workstream program. If the brief is a new platform, a mobile app, and an agent program at the same time, you want a team. If the brief is to put one agent into one operation, or to add one model to the product you already have, one senior engineer is often the whole team you need. I do not run an agency. [Services](https://fractionalappliedai.com/services) are the four kinds of work I take: agent builds, LLM features in an existing product, workflow automation, and technical strategy. If a job needs a specialist I am not, I will say so on the intro call. ## How this differs from a full-time hire A full-time applied AI engineer is the right hire when the work will still be there in a year, when you want someone who accumulates context without an end date, and when you can wait for a search. A senior search often takes a few months. The salary is one part of the cost. You also pay recruiting, benefits, management time, and the ramp before the first production change. A fractional engagement starts sooner and ends on purpose. You are buying a bounded outcome and the context transfer that makes it stick. You are not creating a seat, and you are not hiring a manager. A few months is a poor way to hire someone whose job is to build a department. Some companies use a fractional engineer to bridge a search: ship the first workflow now, and write the role description from the work rather than from a guess. Others use it instead of a search, because the volume of work is real and is not large enough for a salary. Both are reasonable. They are different decisions. The [pricing page](https://fractionalappliedai.com/pricing) separates a monthly retainer from a fixed sprint so you can pick the one that matches. ## When to hire one Hire a forward deployed engineer when most of the following are true. ### You can point at a workflow "We want AI" is a mood. "A dispatcher still copies the same six fields between two systems, and the mistakes show up on the invoice" is a brief. The workflow can live in a product your customers use, or in an internal process your staff run every day. It needs a name, a person who does it today, and a cost when it goes wrong. ### Someone inside the company can make time The engineer cannot discover the real process from a slide. An operator, a founder, or a product lead has to be available for questions and for reviewing the result against a real case. An hour a week of honest review is worth more than a stack of requirements written in the abstract. ### The code has somewhere to land If every change must wait on a vendor's roadmap, an embedded engineer has nowhere to put the work. The best engagements have a staging environment, a way to ship, and a person who can approve a release. If your product is a closed SaaS with no extension point, the engagement may be advisory only: help you decide whether to switch, integrate at the edges, or wait. ### The problem is worth the engagement A model feature that saves a few minutes a week for one person is rarely worth a senior engineer. A loop that ties up several people every day, or a product gap your customers already feel, usually is. You should be able to say, in plain language, what changes in the business if this works. ### You want the system to remain The point of the handover is that your team can run the system after I leave. If the goal is a demo for a single meeting, say that. It is a smaller job, and I will tell you if it is a poor use of a forward deployed engagement. ## When another shape is a better fit Skip this hire when there is no process attached to the request. A chat prototype with no workflow will not survive contact with your data, and you will have paid for a demonstration. Skip the build when you have not chosen a vendor and the real job is the choice. That is advisory work. Building and brokering at the same time muddies who is accountable for the result. Skip it when the missing piece is a product decision you are unwilling to make. An engineer can implement a scope. An engineer cannot pick your market, your pricing, or which customer you serve. Skip it when no one on your side will use the result. A workflow nobody adopts is a failed engagement even when the code is correct. Adoption is part of the scope, which is why the person who does the job has to be in the reviews. ## Fractional cost against a full-time hire The figures on the [pricing page](https://fractionalappliedai.com/pricing) are filled in with the written scope. A number without a scope is a guess, so this page compares the shape of the cost, not a salary table. A full-time senior engineer costs a salary, benefits, taxes, equipment, and a share of management. You pay that in a quiet month and in a busy one. You also pay the search, and you pay the months before the person is useful in your domain. In return you get continuity, and someone who can take a widening brief over years. A fractional monthly retainer costs a fixed amount for a defined slice of time, for as many months as the work justifies. You do not pay benefits or a recruiting fee. You stop paying when the workflow is shipped and handed over, unless you choose to extend. You give up continuity. A person who is with you part of the week, or for one quarter, will not hold the whole company's technical context. That is the trade, and it is a good trade when the work is bounded. A fixed-scope sprint costs a single price for a single outcome. It fits when you can describe the outcome tightly. It fits poorly when the brief is open, because a fixed price around an open brief produces a thin result or a fight about what was included. Advisory hours cost the least and produce a decision, not a production system. Use them to decide. They are a poor substitute for the build. The practical test is the volume of work over the next two or three quarters. If you can name a sequence of workflows that will keep a strong engineer busy for a year, start the full-time search, and use a fractional engineer only if something must ship before that person starts. If you can name one to three workflows and then the need drops, a retainer or a sprint is the better buy. If you cannot name the workflow yet, start with a short advisory block, then decide. On the intro call I will ask which of those three situations you are in. If a full-time hire is the better answer, I will say so. ## How an engagement runs The work runs in three steps. They are the same steps as on the home page, in more detail here. ### Scope the workflow The intro call is 30 minutes. You describe the operation and the outcome you care about. If it is a fit, I follow up with a short written scope: the job, the systems it touches, what is out of scope, and which engagement type matches. You should be able to hand that note to a colleague and have them understand it without a second meeting. ### Embed and ship I join the tools your team already uses. I work in small releases, and I check the result against a real case with the person who owns the process. If you have engineers, I review with them. If you do not, I still write the code so a future engineer can read it. I will not leave you with a private environment only I can run. ### Hand it over Handover is a working path. The feature is in the environment your team uses. Someone other than me has run a real case. The document you already keep, or a short README if you have none, says how to operate it, how to tell when it is wrong, and how to turn it off. After that, the retainer can end or it can take the next workflow. Ending is a normal outcome. ## What you should have at the end You should have a change in production, or a clear written decision to leave the idea alone. You should have the code in your accounts. You should have one person on your team who can explain the system without me in the room. You should know what the model is allowed to do, what a person still checks, and what happens when the model is wrong. Accuracy has to be measured on your cases. A public benchmark says nothing about your invoices, your tickets, or your field notes. Part of the engagement, when we are building, is a small set of real examples and a rule for what "good enough to ship" means on those examples. You should also have a plain account of the risks you accepted. Where the model can write, where it can only draft, and which action still requires a person. That account is part of the handover because the next person to change the system needs it. ## Who this is for I work with founders, CTOs, and operations leaders. The company is small or mid-size. There is a product or an operation with real users. The AI work has a home in that system, and the person hiring wants an embedded engineer for a few months rather than a permanent role. My own background is senior full stack development. I built WellsManager, operations software for oil and gas teams, covering field maintenance, well costs, lease operating statements, and investor payments, with an AI assistant connected to those records. The [about page](https://fractionalappliedai.com/about) has that in one place. If the description on this page matches the gap you have, the next step is a conversation, not a proposal. Use the [contact page](https://fractionalappliedai.com/contact) to book the intro call or to send an email. If you want the offer in list form first, read [services](https://fractionalappliedai.com/services) and then [pricing](https://fractionalappliedai.com/pricing) . ## Questions people ask before they book ### Is a forward deployed engineer the same as a fractional CTO? No. A fractional CTO owns technical direction across the company. A forward deployed engagement is centered on building a specific system with the people who will use it. Advisory hours can cover a slice of the direction. The build is a separate engagement. ### Do you replace our engineers? No. Where you have engineers, I work in their repository and the result should be something they can extend. Where you do not, I still ship, and I will tell you what kind of person you will eventually need. ### Can you work with the stack we already have? That is the job. The usual fit is a web product with an API, a database, and a way to deploy. I will say so on the intro call if the stack is a poor fit, rather than learn it at your expense. ### How do we start? Fill out the short intro call request or send an email. Bring one workflow you wish was less manual, and the name of the system it lives in. You do not need a specification. ```json {"@context":"https://schema.org","@graph":[{"@id":"https://fractionalappliedai.com/#website","@type":"WebSite","description":"Forward deployed engineering for companies that need AI agents and LLM features shipped into the workflows they already run.","name":"Fractional Applied AI","url":"https://fractionalappliedai.com/","publisher":{"@id":"https://fractionalappliedai.com/#identity"}},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#webpage","description":"What a forward deployed engineer does, how the role differs from a consultant or an agency, and when a fractional hire fits.","name":"What is a forward deployed engineer?","url":"https://fractionalappliedai.com/forward-deployed-engineer","@type":["WebPage","FAQPage"],"breadcrumb":{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/breadcrumb-list/#breadcrumb"},"isPartOf":{"@id":"https://fractionalappliedai.com/#website"},"mainEntity":[{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/question/1"},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/question/2"},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/question/3"},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/question/4"}],"potentialAction":[{"@type":"ReadAction","target":["https://fractionalappliedai.com/forward-deployed-engineer"]}],"primaryImageOfPage":{"@id":"https://fractionalappliedai.com/#logo"}},{"@id":"https://fractionalappliedai.com/#identity","@type":"Organization","description":"Hire a fractional applied AI software engineer. AI agents and LLM features shipped into your existing product and workflows.","email":"hello@fractionalappliedai.com","name":"Fractional Applied AI","url":"https://fractionalappliedai.com","contactPoint":{"@type":"ContactPoint","contactType":"sales","email":"hello@fractionalappliedai.com","url":"https://fractionalappliedai.com/contact","availableLanguage":"English"},"logo":{"@id":"https://fractionalappliedai.com/#logo"}},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/breadcrumb-list/#breadcrumb","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","name":"Home","item":"https://fractionalappliedai.com/","position":1},{"@type":"ListItem","name":"Forward deployed engineer","item":"https://fractionalappliedai.com/forward-deployed-engineer","position":2}]},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/question/1","@type":"Question","inLanguage":"en","name":"Is a forward deployed engineer the same as a fractional CTO?","acceptedAnswer":{"@type":"Answer","text":"No. A fractional CTO owns technical direction across the company. A forward deployed engagement is centered on building a specific system with the people who will use it. Advisory hours can cover a slice of the direction. The build is a separate engagement."}},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/question/2","@type":"Question","inLanguage":"en","name":"Do you replace our engineers?","acceptedAnswer":{"@type":"Answer","text":"No. Where you have engineers, I work in their repository and the result should be something they can extend. Where you do not, I still ship, and I will tell you what kind of person you will eventually need."}},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/question/3","@type":"Question","inLanguage":"en","name":"Can you work with the stack we already have?","acceptedAnswer":{"@type":"Answer","text":"That is the job. The usual fit is a web product with an API, a database, and a way to deploy. I will say so on the intro call if the stack is a poor fit, rather than learn it at your expense."}},{"@id":"https://fractionalappliedai.com/forward-deployed-engineer#/schema/question/4","@type":"Question","inLanguage":"en","name":"How do we start?","acceptedAnswer":{"@type":"Answer","text":"Fill out the short intro call request or send an email. Bring one workflow you wish was less manual, and the name of the system it lives in. You do not need a specification."}},{"@id":"https://fractionalappliedai.com/#logo","@type":"ImageObject","caption":"Fractional Applied AI","contentUrl":"https://fractionalappliedai.com/logo.png","inLanguage":"en","url":"https://fractionalappliedai.com/logo.png"}]} ``` ## Sitemap See the full [sitemap](https://fractionalappliedai.com/sitemap.md) for all pages.