AI Government Engineer Job Description

An AI Government Engineer builds AI tools for an agency's own staff and programs, then stays to run them. The work starts with a request from a program office: a benefits team buried in correspondence, an HR shop classifying position descriptions by hand, a policy office whose staff spend hours searching old guidance. The engineer scopes the problem, picks or tunes a model, wires it into agency data and systems, and puts it in front of users. OPM's own AI strategy describes this kind of internal work, using AI to categorize position descriptions and enhance classification systems such as USAClass.

The governance layer shapes every project. OMB memo M-25-21, dated April 3, 2025, sets the rules for how executive agencies use AI. Each agency designates a Chief AI Officer, each CFO Act agency convenes an AI governance board, and every system the engineer builds is screened to decide whether it is high-impact. For high-impact systems, the memo's minimum practices apply: agencies must conduct pre-deployment testing, complete an AI impact assessment before deployment, conduct ongoing monitoring that includes periodic human review where feasible, and ensure human oversight, intervention and accountability. The engineer drafts much of that documentation, since the engineer knows what the model does and where it fails.

A typical portfolio mixes three kinds of work:

• Building. Retrieval assistants over manuals and regulations, classifiers for incoming documents, summarization tools for case files, and code helpers for the agency's own developers. A build might combine Python, an orchestration framework, a vector store and a hosted model behind an agency API.
• Running. Watching accuracy over time, retraining when data shifts, handling user complaints, and keeping each system inside its authority to operate. The information system security officer, privacy officer and records officer are regular collaborators.
• Sharing and reporting. M-25-21 tells agencies to proactively share custom-developed code, including models and model weights, across the federal government, and to reuse what others have built. The engineer packages that code and keeps the agency's annual AI use case inventory entries accurate.

Agencies do not always build from scratch. GSA announced an in-house AI tool for its staff on March 20, 2025, saying it aimed to provide a secure and reliable alternative to AI tools that may not meet the security and privacy requirements necessary for government use. GSA's AI compliance plan describes its USAi platform as a FedRAMP-authorized environment for other agencies to deploy and operate generative AI, with a chatbot interface and API access, so an engineer at a smaller agency may integrate a shared service instead of standing up infrastructure.

The engineer also sits between the agency and its vendors. When the agency buys a model or AI service, the engineer writes the technical requirements with the contracting officer and tests what comes back. The day-to-day audience is internal: program managers, analysts, caseworkers and the governance board, not outside clients.

There is no single federal series for this job. OPM's April 2024 AI classification guidance found AI work spread across 29 occupational series, with the 2210 information technology management series, the 0343 management and program analysis series and the 1560 data science series the most common. The guidance also lists the position classification flysheet for the GS-1550 computer science series as a tool to evaluate positions with AI roles. Which series a posting uses determines the qualification standard an applicant is measured against, so read that section of the announcement first.

As a rough guide, candidates come in with a bachelor's degree in computer science, data science, mathematics, statistics or engineering, and a master's helps for research-heavy positions. Three to five years of software or machine learning work is a reasonable starting point for a journey-level position, and senior positions call for experience owning a production system end to end. Assessment can be hands-on: some of OPM's 2026 cross-government technology pools, for example, used a third-party technical assessment, resume review, and a screening and technical interview.

M-25-21 itself states the profile agencies are told to look for: agencies should focus recruitment efforts on individuals that have demonstrated operational experience in designing, deploying and scaling AI systems in high-impact environments. Portfolio evidence speaks to that directly. A candidate who can describe how they tested a model for uneven error rates, what they monitored after launch, and when they rolled something back is speaking the language of the job.

Core technical skills:

• Python, SQL and at least one deep learning framework such as PyTorch
• Retrieval-augmented generation, embeddings, prompt design and evaluation harnesses
• Cloud platforms used in government (AWS GovCloud, Azure Government, Google Cloud), containers and CI/CD
• Model evaluation: test set design, error analysis by subgroup, drift detection
• Security basics: identity and access management, logging, and working within an authority to operate

Policy fluency is the second half of the job: reading OMB memos, privacy impact assessments, records schedules and accessibility rules, then turning them into design decisions.

Certifications are optional but useful signals:

• AWS Certified Machine Learning Engineer - Associate (Amazon Web Services)
• Professional Machine Learning Engineer (Google Cloud)
• Artificial Intelligence Governance Professional, AIGP (International Association of Privacy Professionals)
• CompTIA Security+ (CompTIA)

Expect a background investigation before starting, and for defense, intelligence or homeland security work, possibly a security clearance. Plan to document foreign contacts and travel, and to wait while the investigation runs.

BLS does not count AI government engineers separately. The closest occupation, computer and information research scientists, had 37,200 jobs nationally in May 2025, and the federal executive branch employed 11,030 of them.

Reported AI use inside agencies has grown. GAO found that across 11 selected agencies with AI inventories, reported AI use cases nearly doubled from 571 in 2023 to 1,110 in 2024, while generative AI use cases rose from 32 to 282. OMB's consolidated 2025 inventory, summarized as of April 13, 2026, counted 3,611 individually reported AI use cases at all stages of development, 445 of them high-impact. Each of those entries is a system someone has to build or buy, monitor and report on.

The policy frame was still in place in 2026. M-25-21 required agencies to document implementation of the minimum practices for high-impact AI within 365 days of its issuance, and a Department of Energy acquisition letter dated May 8, 2026 still relied on it. That same letter describes OMB memo M-26-04, issued December 11, 2025, which established two Unbiased AI Principles, truth-seeking and ideological neutrality, that agencies must enforce when buying large language models, and agencies are required to include the principles in new solicitations and orders for them.

Shared hiring pools are another route in. In its June 2026 summary of cross-government hiring actions, OPM reported 3,529 qualified candidates ready for agencies to hire, 1,829 of them in digital, IT and tech fields, a count that includes U.S. Tech Force candidates. OPM says these shared hiring actions reduce duplicative recruiting and accelerate selections.

Agencies report real constraints. GAO wrote that agency officials face several challenges to using generative AI, such as complying with existing federal policies and guidance and having sufficient technical resources. Eight selected agencies, including Commerce, DHS, DoD and VA, reported challenges in obtaining or accessing the needed technical resources. GSA's compliance plan says USAi lets agencies integrate AI into mission workflows without the need to build their own infrastructure.

At the state level, NASCIO's 2026 State CIO Top 10 Priorities list puts artificial intelligence, generative AI, agentic AI and machine learning at number one, covering governance, use cases, security, privacy and workforce skills.

Career paths run in two directions. Technical tracks lead to senior and lead AI engineer roles, then to architect positions responsible for an agency's shared AI platform. Governance tracks lead toward deputy Chief AI Officer and AI program manager roles, where the engineer's knowledge of how systems fail informs policy. Some engineers move between agencies through shared hiring pools; others move to state government or to contractors serving the same agencies.

Dear Hiring Manager,

I am applying for the IT Specialist (Artificial Intelligence) position in your Office of the Chief Information Officer. For the past four years I have built and run machine learning systems for a state workforce agency, and I want to bring that experience to a federal team that owns its tools for the long term.

My main project was a document classifier for unemployment insurance correspondence. Claimants sent letters, appeals and identity documents into one inbox, and staff sorted them by hand. I built the classifier, wrote the test plan with our program staff, and set up a weekly review where an adjudicator checked a sample of the model's routing decisions. When a change in the claim form shifted the input data, our monitoring caught the drop in accuracy and I retrained the model before the backlog grew. I also wrote the plain-language description of the system that went to our legislature's oversight committee.

I have read OMB memo M-25-21 closely, and the minimum practices for high-impact AI match how I already work: test before launch, document the risks, monitor after launch, and keep a person accountable for every decision that affects a member of the public. I would rather spend a week writing an impact assessment than ship a tool I cannot explain.

On the technical side, I work in Python and PyTorch, have built retrieval assistants on a cloud government region, and have worked with our security office to keep systems inside their authority to operate. I hold the AWS Certified Machine Learning Engineer - Associate credential.

What draws me to your office is the chance to build something and stay with it: to own the retraining, the inventory entry and the hard decision to retire a model when it stops serving users. I would welcome the chance to talk about how I can help your team.

Sincerely, Jordan Ellis