These are build-and-ship jobs. So why do they demand a Computer Science degree?
I took ten real forward-deployed, AI-implementation, and software-builder postings and rewrote them to ask for what the job actually needs - then showed my work.
The big idea: A job posting is supposed to describe a job. But most of them describe a shopping list - a degree, a pile of years, and every tool anyone in the room has ever heard of. Then an automated filter throws out the people who could actually do the work. Let's fix that.
Here's the thing. The hottest roles in tech right now are applied. Forward deployed engineer. AI implementation engineer. Solutions engineer. The person who drops into a customer's mess, wires up an LLM, and ships something that works on Friday. And I keep seeing these exact roles gated behind a Computer Science degree, or "5+ years of a framework that's barely three," or a wall of twenty tools.
None of that tells you whether the person can do the job. It's a proxy. And the machine screening resumes doesn't know it's a proxy - it just rejects anyone who doesn't match the keywords.
To be fair: if the job is inventing machine learning algorithms or writing models from scratch, a computer science background earns its place. Sure. But that is not most of these jobs. A forward deployed engineer, an AI implementer, a builder wiring LLMs into a product - that is a build-and-ship job, and a CS degree was never the thing that made someone good at it.
The people you lose by gating on the degree are exactly the ones you want right now: the self-taught builder, the career-changer, the person who learned by shipping instead of by sitting in a lecture hall.
Even Boris Cherny, the guy who created Claude Code, doesn't have a traditional CS degree. Asked about it, his advice was simple: learn by solving real problems. That's the whole game now.
"The world is changing at such a rapid rate that it's turning us all into amateurs. Even for professionals, the best way to flourish is to retain an amateur's spirit and embrace uncertainty and the unknown."
Austin Kleon, Show Your Work! (2014)
If everything's changing this fast, the smartest thing you can hire for isn't a credential someone earned a decade ago. It's a track record of learning and adapting. So hire the people who have already proven they can change with the change.
What the rewrite actually does
Every posting below went through the same skill. It does four things:
Figures out what the person will actually do - the real problems and outcomes, not a keyword list.
Strips the proxies: degree gates that aren't legally required, "X years of [tool]," and tool bingo.
Replaces them with the real signal - capabilities tied to the work, and evidence of a track record.
Opens the door to non-traditional paths, so a strong builder never gets filtered out before a human reads them.
The bar doesn't drop. It moves onto the actual job. Here's what that looks like on ten real postings, live as of August 2026. Each one links to the original so you can see it for yourself.
Ten real postings, rewritten
Forward deployed engineers, AI implementers, solutions engineers, software builders. Left is the posting as written, quoted from the source. Right is the same role, rewritten to hire for the work.
"Bachelor's or Master's degree in Computer Science, Engineering, or related discipline"
"5+ years in software development, systems integration, or platform engineering"
TypeScript/JavaScript/Python, API integration, serverless deploys, LLM/RAG familiarity
Rewritten to hire for the work
What you'll actually do
Deploy DevRev into real customer environments: integrate their systems, sync data across platforms, and wire up LLM-powered workflows that hold up in production.
What you need to be able to do
Strong TypeScript/JavaScript and Python.
You've integrated real systems via APIs, webhooks, and data pipelines.
Comfortable deploying on serverless/edge (Lambda, Cloud Functions).
Practical grasp of LLMs, RAG, and function calling - built with them.
Nice to have
Large-scale data synchronization; schema mapping across messy, heterogeneous systems.
How to show us you're a fit
An integration or app you shipped for a real user. This is a build-and-ship role - show us what you've built, degree or no degree.
Why it changed: the work is integrating systems and shipping LLM workflows into customer environments. A CS degree filters out the exact builders who are best at that.
"Bachelor's or Master's degree in Computer Science, Engineering, or a related field."
"3+ years ... software development/consulting, AI/ML engineering, or system integration, or as FDE"
Python and Golang; LLMs, prompt engineering, agent frameworks, RAG; APIs/webhooks/pipelines
Rewritten to hire for the work
What you'll actually do
Stand up Cresta's AI agents inside real customer operations: integrate their systems, tune prompts and agent workflows, and make the thing perform in production.
What you need to be able to do
Solid Python (Golang a plus).
Hands-on with LLMs, prompt engineering, agent frameworks, and RAG.
You've integrated systems via APIs, webhooks, and pipelines.
Comfortable with a cloud platform and basic DevOps.
Nice to have
Contact-center or CX domain experience; A/B testing and performance monitoring.
How to show us you're a fit
An AI agent, integration, or automation you've built and shipped for real users.
Why it changed: deploying AI agents into customer operations is integration and prompt/agent work. That's a build skill, not a diploma - so we asked for evidence of building it.
Rewritten with the rewrite-job-requirements skill
Senior Forward Deployed Engineer
Liberate Innovations · Boston, MA / SF (Hybrid) · Full-time
"5+ years of experience in a solutions engineering or technical consulting role..."
"Bachelor's degree in Computer Science, Engineering, or a related field"
Python/Java, APIs, data integration, cloud platforms
Rewritten to hire for the work
What you'll actually do
Own customer deployments end to end for an early-stage AI company: understand the problem, integrate Liberate into their stack, and tell the technical story clearly to both engineers and executives.
What you need to be able to do
A track record in solutions engineering or technical consulting, ideally in AI/tech.
Comfortable in Python or similar, with APIs, data integration, and cloud.
Explain complex things simply to any audience.
Juggle multiple customer engagements on tight timelines.
Nice to have
Insurance or financial-services domain; cloud certs; AI/ML familiarity.
How to show us you're a fit
Customer deployments or integrations you've owned.
Why it changed: this is a solutions and consulting role - talk to customers, integrate, present. Requiring a CS degree for it is the clearest "why?" on the page. A diploma was never what made someone good at this.
"Undergraduate/graduate degree in Computer Science, or related degree - degree must be completed before joining"
Go/Python/Java/JS/C++; data structures, algorithms; distributed systems; shipping ML/NLP
Rewritten to hire for the work
What you'll actually do
Build and ship features across Glean's product and infrastructure, working alongside experienced engineers on real, customer-facing systems.
What you need to be able to do
Strong coding in a modern language (Go, Python, Java, JavaScript, or C++).
Solid grasp of data structures, algorithms, and clean software design.
Curiosity and a fast learning curve.
Nice to have
Exposure to distributed systems or cloud-native apps; ML or NLP projects.
How to show us you're a fit
Things you've built: projects, a portfolio, a repo, internship work, open source.
Why it changed: "degree must be completed before joining" has no equivalent-experience path - it auto-rejects the self-taught grad and the bootcamper who can already code. We asked to see what they've built instead.
Deploy AI/ML solutions into real customer environments end to end: design, build, validate, and get them running reliably on the customer's infrastructure.
What you need to be able to do
Real end-to-end AI/ML project experience: design, build, validate.
Strong Python.
Comfortable deploying into enterprise infrastructure and cloud.
Can integrate data from messy, real-world sources.
Nice to have
Data-center tech (OpenShift, VMWare, Linux, networking); specific ML frameworks; IBM WatsonX. Helpful context, not a checklist to match line for line.
How to show us you're a fit
An AI/ML solution you took from design to running in someone's environment.
Why it changed: the original named about fifteen technologies - three jobs stapled together. We kept the few that carry the role and moved the rest to "helpful context," so a great implementer isn't rejected for never touching Scribus.
"3-5 years of full-stack software engineering experience"
"Expert-level programming skills in at least two languages"
"A bachelor's degree in Computer Science, ECE, Engineering, Information Systems, or equivalent practical experience" (the good part)
Rewritten to hire for the work
What you'll actually do
Drop into customer environments and rapidly build production-ready apps on Superblocks: prototype fast, integrate their systems, and ship something real under ambiguity.
What you need to be able to do
Strong full-stack skills; you can prototype fast and harden apps for production.
Comfortable in Python and JavaScript/TypeScript.
Good with modern cloud architectures; you thrive in fast, ambiguous, customer-facing settings.
Communicate clearly with technical and non-technical people.
Nice to have
API design and integration; enterprise security; data pipelines.
How to show us you're a fit
Apps and prototypes you've shipped, especially anything you built fast in front of a customer.
Why it changed: credit to Superblocks for allowing "equivalent practical experience" on the degree - that's the right call. We just reframed "expert-level in at least two languages" and the year range as the ability the job needs: can you build and ship fast?
"Bachelor's degree in Computer Science, Engineering, or a related field (or equivalent practical experience)"
"Strong foundations in computer science and software engineering principles"
A stacked list: AI-native apps, RAG, agent orchestration, frontend system design, backend/infra, production LLM systems
Rewritten to hire for the work
What you'll actually do
Build AI-native applications in and for customer environments: chat and agent interfaces, RAG systems, the full stack from frontend to the LLM pipeline.
What you need to be able to do
You've built AI-native apps (chat/agent interfaces, RAG).
Real production LLM experience: RAG pipelines, agent orchestration, embeddings/vector search, prompt design.
Solid full-stack instincts, frontend through backend, APIs, and auth.
You can own a system end to end, from problem definition to production.
Nice to have
Client-facing or high-ownership environments.
How to show us you're a fit
An AI-native app you built and shipped - the repo, the demo, the story of what you owned.
Why it changed: they already offered "equivalent practical experience" - good. The job is building AI apps end to end, so we asked for evidence of exactly that instead of a "strong CS foundations" abstraction.
Then ten stacked "Experience with..." lines: 5+ years, a language, GenAI in production, technical leadership, REST/microservices/distributed systems, cloud, CI/CD and DevOps, customer-facing communication, leading teams, and travel.
Rewritten to hire for the work
What you'll actually do
Lead customer deployments of AI/GenAI solutions end to end: translate a business problem into a working system, build it, integrate it, and ship it in the customer's environment.
What you need to be able to do
You've built and deployed real software and AI/GenAI solutions in production.
Comfortable across a modern language and a cloud platform.
You can lead a technical workstream and talk directly to customers.
Nice to have
REST/microservices/distributed systems, CI/CD and DevOps, prior team leadership. Strengths, not gates.
Required (this one's real)
US Citizen or Permanent Resident able to obtain Public Trust clearance. A genuine requirement, so it stays.
How to show us you're a fit
An AI solution you led from business problem to production.
Why it changed: ten "must have experience with X" lines describe a whole team, and read as "reject anyone missing any single line." We kept the few that define the role and turned the rest into strengths.
An AI thing you've actually built - a RAG app, an LLM feature, a demo repo.
Why it changed: the title says "All Levels," then the requirements gate on a degree and 2-5 years. For an applied builder role, we kept the one question that matters at every level: can you build this and ship it?
"A graduation date in Fall 2025 or Spring 2026 with a Bachelor's degree (or equivalent) in a relevant field (Computer Science, EECS, Computer Engineering, Statistics)"
Build product full-stack: ship web apps, wire up APIs and services, talk to customers, and figure out what to build, then build it.
What you need to be able to do
You can build full-stack and integrate real APIs and services.
Solid coding in something like Python, TypeScript, or React.
A track record of shipping - from internships, projects, or your own builds.
Nice to have
Experience with systems that process large volumes of data; MongoDB or similar.
How to show us you're a fit
Products and features you've shipped: a portfolio, a repo, internship work, a side project.
Why it changed: credit for the "(or equivalent)" - but listing only CS-family majors as "relevant" still signals what they're really screening for. The work is building and shipping product, so we asked for evidence of shipping.
Rewritten with the rewrite-job-requirements skill
Get the skill
Just save this skill to your AI skills and start updating your copy.
rewrite-job-requirements / SKILL.md
---
name: rewrite-job-requirements
description: >-
Rewrites a bloated, buzzword-heavy, or credential-gated job posting into a
clear posting built around what the role actually needs to get done. Use this
whenever someone shares a job description, a requirements list, a "must-haves"
bullet list, or a JD and wants help fixing it, tightening it, or making it
attract the right people. Trigger it even when they do not say the word
"rewrite" - phrases like "does this job post look right?", "help me write
requirements for this role", "why aren't we getting good candidates",
"clean up this JD", "what should this role actually require", or pasting a
posting with a vague ask all count. Especially use it when the posting leans
on a degree requirement, a long laundry list of tools, or "X years of
[buzzword]", because that is exactly the problem this skill fixes.
---
# Rewrite Job Requirements
## What this skill is for
Most job postings are written by pasting together every tool, credential, and
buzzword anyone in the room has ever heard of. The result is a wish list for a
person who does not exist, screened by an automated filter that rejects the
people who could actually do the work. The classic tells: a required degree
that has nothing to do with the daily job, "8+ years of [framework that is 6
years old]", and thirty bullet points that describe an entire department rather
than one role.
Your job is to turn that into a posting built around one question: **what does
this person actually need to accomplish, and what is real evidence they can do
it?** You strip the proxies (degree, years, tool-name bingo) and replace them
with the real thing (the problems to solve, the outcomes, and the kind of track
record that predicts success).
The guiding belief, and it is worth saying out loud to the user if it helps:
the world is changing fast enough that everyone is a bit of an amateur now. The
best hire is usually someone with a track record of learning and adapting, not
someone who collected the right credentials a decade ago. Write the posting to
find that person.
## The method
Work in this order. Do not jump straight to prose.
### 1. Figure out what the job actually is
You cannot write real requirements from a buzzword list, because the buzzword
list does not tell you what the person will do on a Tuesday. Read what the user
gave you and extract, or ask for, these things:
- The 3 to 5 problems this role exists to solve, or outcomes it owns.
- What a normal week actually looks like (what they build, decide, ship, or fix).
- Who they work with and what "good" looks like after 6 to 12 months.
- The hard constraints that are genuinely non-negotiable and why.
If the posting is all buzzwords and you cannot tell what the job is, do not
invent it. Ask the user a few pointed questions first: "What will this person
actually spend most of their time doing?" and "If they nail this role, what is
different in a year?" A short, honest set of answers beats a beautiful posting
built on guesses. Never fabricate specifics about the team, salary, stack, or
mission that the user did not give you.
### 2. Separate real requirements from proxies
A **real requirement** is something the person must be able to do on day one or
learn very fast, tied directly to the work. A **proxy** is a stand-in someone
reached for because it felt safer to filter on. Proxies are where good
candidates get thrown out by a machine before a human ever reads them.
Challenge these hard. For each one, either cut it or convert it into the real
signal underneath:
- **Degree requirements.** Unless a license or accreditation is legally
required (nurse, PE, CPA, etc.), a degree is almost always a proxy for "can
think, learn, and do the work." Cut "Bachelor's in Computer Science required"
and replace it with the actual capability and how a candidate can show it
(shipped projects, a portfolio, a track record). If the user insists a degree
matters, push once and ask what it is really standing in for, then capture
that instead.
- **"X years of [tool/skill]."** Years are a weak proxy for competence and a
hard gate that filters out fast learners and career-changers. Replace "5+
years of React" with the outcome ("comfortable owning a production frontend
and shipping features independently"). Keep a seniority signal only where it
genuinely matters (for example, "has led a team through a real migration"),
and phrase it as the experience, not the stopwatch.
- **Tool laundry lists.** Ten named technologies usually means nobody decided
which ones matter. Keep the two or three that are actually load-bearing, and
fold the rest into "or similar" so you are hiring for the underlying skill,
not the exact logo.
- **Culture-fit and vague adjectives.** "Rockstar", "ninja", "works well under
pressure", "self-starter" screen for nothing and quietly filter for bias.
Replace with a concrete behavior tied to the work ("comfortable making calls
with incomplete information and adjusting as you learn").
The point is not to lower the bar. It is to move the bar to something real, so
the filter keeps the people who can do the job instead of the people who happen
to match keywords.
### 3. Prioritize evidence of a track record
Because the strongest predictor of thriving amid change is having done it
before, ask explicitly for evidence, not credentials. Invite candidates to
point at things they have built, shipped, fixed, or figured out - a repo, a
portfolio, a project write-up, a story of a hard problem they owned. Make it
clear that self-taught, career-changer, and non-traditional paths are welcome
if the track record is there. This is the single change that most widens the
pool toward people who actually deliver.
### 4. Write the reworked posting
Now write it, using the template below. Keep it tight. A shorter posting that
is honest about the work will out-recruit a long one full of hedges.
## Output template
Produce a full reworked job post in this shape. Adapt headings to fit the role,
but keep the spirit: lead with the work, make must-haves real and few, and make
the application path welcoming to non-traditional backgrounds.
```
# [Role title]
## What you'll actually do
2 to 4 sentences on the real work and the problems this role owns. Concrete,
not aspirational. A candidate should picture their Tuesday.
## What success looks like
3 to 5 bullets describing outcomes in the first 6 to 12 months. Written as
results, not activities.
## What you need to be able to do
The real must-haves - capabilities tied to the work, kept short (aim for 4 to
6). No degrees-as-gates, no year counts unless they genuinely matter, no tool
bingo. Phrase each as a demonstrable ability.
## Nice to have
Things that would help but are not gates. This is where "familiarity with X"
and bonus experience live, so they never block a strong candidate.
## How to show us you're a fit
Invite evidence of a track record: projects, a portfolio, a repo, a short note
on a hard problem they solved. Explicitly welcome self-taught and
non-traditional paths.
## [Optional: about the team / logistics]
Only include facts the user actually provided (comp, location, stack, mission).
Do not invent these.
```
## Before / after example
**Before (what the user pastes in):**
> Senior Software Engineer
> - Bachelor's degree in Computer Science or related field required
> - 8+ years of professional software development experience
> - 5+ years of React, 5+ years of Node.js, experience with AWS, Docker,
> Kubernetes, GraphQL, PostgreSQL, Redis, Kafka, Terraform
> - Strong communication skills and ability to work in a fast-paced environment
> - Self-starter who can wear many hats
> - Rockstar engineer who thrives under pressure
**After:**
> ## Senior Software Engineer
>
> ### What you'll actually do
> Own major features on our customer-facing web app end to end, from a rough
> idea to something running in production. You will make real architecture
> calls, keep the system reliable as we grow, and help newer engineers level
> up.
>
> ### What success looks like
> - In 6 months, you have shipped at least one significant feature that you
> scoped, built, and are now maintaining.
> - You are the person the team trusts to untangle a gnarly production issue.
> - Your reviews and decisions have made the codebase easier for everyone.
>
> ### What you need to be able to do
> - Ship production web features independently, front to back.
> - Reason about a live system: performance, reliability, and what breaks at
> scale.
> - Work comfortably in a JavaScript/TypeScript stack (we use React and Node;
> related experience transfers fine).
> - Communicate a technical tradeoff clearly enough that a teammate can act on it.
> - Make good calls with incomplete information and adjust as you learn.
>
> ### Nice to have
> Experience with cloud infrastructure (we run on AWS), containerized
> deployments, or GraphQL. Helpful, not required - you can pick these up here.
>
> ### How to show us you're a fit
> Point us at something you have built or shipped: a repo, a portfolio, a
> project, or a short note on a hard problem you owned and how you worked
> through it. We care about your track record, not your diploma. Self-taught
> and non-traditional paths are genuinely welcome.
Notice what changed: the degree gate is gone, the year counts became
capabilities, nine tools collapsed into "the stack plus related experience
transfers", and the empty adjectives became one concrete behavior. The bar did
not drop. It moved onto the actual work.
## Guardrails
- **Do not invent facts.** Never make up salary, location, benefits, team size,
or mission. If the user did not provide it, either leave a clearly marked
placeholder or ask.
- **Keep legally required credentials.** Some roles genuinely require a license,
clearance, or accreditation. Keep those. The target is fake proxies, not real
legal requirements.
- **Push, then respect the user's call.** If the user insists on keeping a
degree requirement or a specific year count, challenge it once by asking what
it is really standing in for, then defer to them. It is their hire.
- **Match their voice and constraints.** If the user has a house style, comp
band, or format they must use, work within it. The goal is a better version
of their posting, not a different company's posting.
- **Bias awareness.** Flag language that tends to narrow the pool without adding
signal (gendered terms, "culture fit", age-coded phrases like "digital
native", unnecessary physical requirements) and offer neutral replacements.