AI engineer is a recent title and job adverts using it vary widely. What they share is a product built on a model somebody else trained: retrieval, prompting, tool use, agents, and the evaluation harness that stops the whole thing regressing silently.
Because the title is new, the CV has to do more work to establish credibility. The most persuasive material is evaluation: how you measured whether the system was getting better, what your test set looked like, and how you caught a regression before users did. Anyone can wire up an API call; far fewer can tell you whether the output improved.
The second differentiator is cost and latency discipline. Token spend, caching, model routing and context management are the operational realities of the role.
This example is written for an engineer with four years of software experience, the last two spent building language model features in production.
AI engineer with four years of software experience, the last two building language model features in production. Recent work covers a grounded support assistant used by 60 agents, an evaluation harness that raised grounded-answer rate from 71% to 89%, and a 46% reduction in monthly model spend.
Professional Experience
AI EngineerJune 2023 – Present
Cardamom Support Platform · Bengaluru, India
•Built the support answer assistant used by 60 agents, grounded on 12,000 help-centre documents with hybrid search and a reranker.
•Created the evaluation harness: a 300-question set graded by two support leads, run on every prompt or retrieval change, with a regression gate in CI.
•Raised grounded-answer rate from 71% to 89% by chunking on document headings rather than fixed length and adding a citation requirement to the output schema.
•Cut monthly model spend by 46% by routing classification calls to a smaller model, caching system prompts and truncating retrieved context above a measured relevance threshold.
•Added a PII redaction step and a human review path for any answer the model marked low confidence, agreed with the legal team before launch.
TypeScript · Python · pgvector · FastAPI
Software EngineerAugust 2021 – May 2023
Cardamom Support Platform · Bengaluru, India
•Built the ticket routing service and the search backend that later became the retrieval layer for the assistant.
•Introduced structured logging and tracing across three services, which made cross-service latency problems diagnosable.
TypeScript · PostgreSQL · Elasticsearch
Selected Projects
evalkit
Author
•A small open-source harness for running graded evaluation sets against language model applications, with regression reporting between runs.
•Extracted from internal use after two other teams asked to adopt it.
Education
B.Tech Computer Science and EngineeringJuly 2017 – May 2021
National Institute of Technology Calicut · Kozhikode, India
First Class with Distinction
AI Engineer example on the Technical layout. All details are fictional and shown for demonstration only.
What recruiters expect
Before writing anything, it helps to know what the person reading is checking for. In this field that is usually a short, specific list:
An evaluation approach: test sets, graders, regression detection - not just vibes.
Retrieval detail: chunking strategy, embedding choice, reranking, and how you measured retrieval quality separately from generation quality.
Cost and latency management, including caching and model selection by task.
Safety and failure handling: refusals, hallucination mitigation, human review paths.
Ordinary software engineering. Most AI features fail on plumbing, not on prompts.
Recommended CV structure
This is the running order the example uses. It is a starting point rather than a rule, but the order reflects what tends to be read first in this profession.
Profile — Three or four lines positioning you for the role.
Technical Skills — Grouped skills, for example "Languages" and "Tooling".
Professional Experience — Paid roles, in reverse chronological order.
Selected Projects — Work you built, with outcomes and the stack used.
Education — Degrees, diplomas and school-leaving qualifications.
Certifications — Completed certifications with the issuing body.
Languages — Spoken languages with CEFR levels.
Sections worth adding
Writing & Talks — The field moves quickly; published write-ups are treated as evidence of depth.
Open Source — Contributions to evaluation or retrieval tooling stand out.
Skills worth including
Grouped rather than listed in one block. Grouping makes a long list readable and shows that you can tell the difference between the things you use daily and the things you have touched.
Beyond the technical list: Managing expectations about model capability, Working with subject-matter reviewers, Writing product-facing documentation. These belong inside your experience bullets, demonstrated, rather than in a list of adjectives.
Example professional summary
Three or four lines, positioned for the role rather than describing your personality. Two versions you can adapt:
AI engineer with four years of software experience, the last two building language model features in production. Recent work covers a grounded support assistant used by 60 agents, an evaluation harness that raised grounded-answer rate from 71% to 89%, and a 46% reduction in monthly model spend.
Engineer working on retrieval and evaluation for language model products. More interested in whether the output is right than in which model produced it.
Writing your experience
The difference between a CV that gets a call and one that does not is almost always in the bullet points. Each pair below shows a real rewrite of the kind of line that appears on most CVs in this field.
Weak
Built a chatbot using an LLM API.
Stronger
Built the support answer assistant used by 60 agents, grounded on 12,000 help-centre documents with hybrid search and a reranker; measured answer usefulness on a 300-question set graded by two support leads.
Names the audience, the corpus, the retrieval design and how quality was actually measured.
Weak
Improved response quality with prompt engineering.
Stronger
Raised grounded-answer rate from 71% to 89% on the evaluation set by chunking on document headings rather than fixed length and adding a citation requirement to the output schema.
A measured improvement attributed to two specific design changes, not to prompt tinkering.
Weak
Reduced API costs.
Stronger
Cut monthly model spend by 46% by routing classification calls to a smaller model, caching system prompts and truncating retrieved context above a measured relevance threshold.
Three concrete levers, in the order of impact.
Taken from the example
The sample CV for this profession is fully written. A few sections from it, so you can see the level of specificity that works:
Experience
AI Engineer, Cardamom Support Platform
Built the support answer assistant used by 60 agents, grounded on 12,000 help-centre documents with hybrid search and a reranker.
Created the evaluation harness: a 300-question set graded by two support leads, run on every prompt or retrieval change, with a regression gate in CI.
Raised grounded-answer rate from 71% to 89% by chunking on document headings rather than fixed length and adding a citation requirement to the output schema.
Cut monthly model spend by 46% by routing classification calls to a smaller model, caching system prompts and truncating retrieved context above a measured relevance threshold.
Projects
evalkit — A small open-source harness for running graded evaluation sets against language model applications, with regression reporting between runs.
Education
B.Tech Computer Science and Engineering, National Institute of Technology Calicut — First Class with Distinction
Certifications and registration
Deep Learning Specialisation — DeepLearning.AI
Common mistakes
Prompt engineering as the headline skill
It is a component, not a discipline. Leading with it suggests you have built demos rather than systems that stayed working.
No evaluation story
This is the single most common gap and the easiest way to stand out. Describe your test set, your grader and one regression you caught.
Ignoring cost
Token spend is a line item finance teams now ask about. A concrete reduction is one of the strongest bullets you can write.
Overclaiming autonomy
"Built an autonomous agent" invites scepticism. Describe the tools it could call, the guardrails, and what a human still checked.
ATS considerations
Applicant tracking systems behave differently by sector, and generic advice is often wrong for a given field. These points are specific to ai engineer applications:
Spell out "retrieval-augmented generation (RAG)" so both the phrase and the abbreviation match.
Write "large language model (LLM)" once in full for the same reason.
Keep conventional software keywords on the CV. Many AI engineer roles are screened by a platform team first.
Name the vector store or search engine specifically rather than writing "vector database".
The Minimal ATS layout is built for this, and the ATS guide covers what parsers do to a file in more detail.
Questions about ai engineer CVs
Do I need a machine learning background?
Rarely for application roles. Strong software engineering plus genuine evaluation discipline is the usual profile. ML depth matters more for fine-tuning and research positions.
How do I show experience when the field is only a few years old?
Depth beats duration here. One production feature with a real evaluation harness is worth more than a year of experiments.
Should I list specific model names?
List the providers and capabilities rather than chasing version numbers, which date quickly. Mention specific models only where the choice was part of the engineering.
Is prompt engineering a job title worth using?
Not on its own in most markets. Position yourself as an engineer who builds AI features; prompting is one of the tools you use.