AI Engineer

JOB DESCRIPTION

Entry Level AI Engineer

Caava Group · Group Technology · Evolve Platform

Reports To

Technical Lead of the assigned Evolve team

Works With

Software Engineers, QA Engineers and Product across Evolve teams

Scope

AI-enabled features across Evolve: data preparation, model/LLM integration, evaluation and monitoring support

Tech Stack

Python, REST APIs, model and LLM APIs, Git and evaluation tools; integration with Evolve’s Spring Boot services

Growth Path

Associate AI Engineer, based on performance and business need

Role Purpose

Evolve is Caava Group’s insurance management platform, built on Spring Boot services with Angular and Next.js frontends, and used by AIICO for its insurance operations. This role helps build, test and maintain AI capabilities used across Evolve teams. Working with experienced engineers, you will use Python to prepare data, integrate approved model and LLM APIs with Evolve services, and evaluate whether AI features behave reliably for their intended use.

This entry-level role focuses on practical implementation and evaluation. It suits candidates with internship, academic or personal project experience who learn quickly, document their work and ask for guidance when needed.

Career Path

Progression to Associate AI Engineer depends on consistent delivery against these measures, growing independence on larger tasks, code and test quality, and business need.

Core Accountabilities

Build AI components and prepare data

• Implement clearly scoped Python components, scripts and integrations to agreed designs, exposed to Evolve services through agreed APIs. Add input validation, error handling and tests with guidance.
• Prepare and validate approved datasets for development and evaluation. Check formats, missing values and labelling quality; document transformations and limitations.

Integrate models and evaluate outputs

• Integrate model or LLM APIs into Evolve workflows. Support prompt configuration, structured outputs and response validation using approved tools and data.
• Create and run repeatable evaluation cases covering typical, edge and failure scenarios. Compare outputs against agreed criteria and document errors, limitations and regressions.

Investigate failures and handle data safely

• Help investigate unreliable outputs and integration failures. Record findings and escalate quality, safety or data concerns to experienced engineers.
• Apply approved privacy and security practices. Keep personal data, credentials and confidential information out of unapproved prompts, datasets, logs and external services.

Support monitoring and engineering practice

• Support deployment and monitoring under guidance. Check API failures, response time, usage and cost, and help implement agreed fallback or human-review steps.
• Use Git, tests and code reviews to keep code readable. Document APIs and evaluation results, apply feedback and communicate progress and blockers early.

Key Performance Indicators

KPI

Measure

Target

Sprint delivery

Original sprint items Done by sprint end ÷ items committed at sprint start × 100.

≥85% each sprint.

Evaluation coverage

Agreed evaluation cases executed and reported ÷ agreed cases × 100.

100% before release; Proposed: no unexplained regression against the last accepted baseline.

Safe data handling

Datasets and integrations with completed handling checks before use ÷ those used × 100.

100%; 0 unapproved data exposures.

Integration reliability

Failed calls, p95 response time and cost for assigned integrations.

Proposed: ≤2% failed calls after retries; p95 response time and monthly cost within agreed limits; fallback tested before release.

Release quality

Own released changes linked to a production defect or rollback ÷ own released changes × 100.

Proposed: ≤10% over a rolling 4 weeks; 0 critical defects.

Code review

Time to first response on assigned reviews.

<6 working hours.

Review weekly with your lead: sprint measures are scored per sprint and release quality over a rolling 4 weeks. Snapshot scope at sprint planning and record changes separately. Report N/A where no work is eligible. Proposed targets are confirmed after the first quarter of data.

Experience and Working Style

• A degree in Computer Science, Software Engineering, Data Science or a related field, or equivalent practical training supported by relevant projects.
• Foundational Python skills shown through coursework, an internship or a working project.
• Basic understanding of machine learning concepts, evaluation metrics and the limits of model outputs, including inaccurate or unsupported responses.
• Familiarity with REST APIs, HTTP and JSON, Git, automated tests and debugging in an existing codebase.
• Ability to work with structured data, use documentation and explain the approach and results of a technical task.

Working Style

• Attention to detail and sound judgement about handling data.
• Clear communication and a willingness to seek and apply feedback.

Nice to Have

• Exposure to FastAPI or Flask, and Python data tools such as pandas or NumPy.
• A project using a model API, LLM, scikit-learn, PyTorch or TensorFlow; experience with every framework is not expected.
• Familiarity with SQL, Docker, Java/Spring Boot, cloud concepts or CI workflows, and interest in practical insurance uses of AI.