AI engineer

Candidate should have the fundamentals to develop and maintain a local AI system.
Python: As the de facto language for AI and machine learning, a strong grasp of Python is essential. This includes:

• Core Python syntax and data structures
• Object-Oriented Programming (OOP) concepts
• Working with libraries like NumPy and Pandas for data manipulation and analysis
• Understanding virtual environments and package management (pip)

Generative AI Frameworks and Libraries

LangChain: This framework is crucial for developing applications powered by large language models. It enables chaining together different components to build more complex use cases, such as:

• Prompt management and optimization
• Retrieval-Augmented Generation (RAG) pipelines
• Agent development and tool integration
• Memory management for conversational AI

PyTorch / TensorFlow: While not strictly necessary for all generative AI roles, understanding these deep learning frameworks allows for deeper customization, model development, and research.

Other Relevant Tools and Concepts

Vector Databases / Vector Stores: Essential for efficient storage and retrieval of embeddings in RAG pipelines.

Examples:

• Pinecone
• ChromaDB
• Weaviate

Cloud Platforms (AWS, Azure, GCP): Familiarity with cloud services for deploying and scaling AI applications.

Examples:

• Amazon SageMaker
• Azure Machine Learning
• Google Vertex AI

MLOps (Machine Learning Operations): Understanding practices for deploying and maintaining machine learning models in production, including:

• Version control
• Monitoring
• Continuous Integration / Continuous Delivery (CI/CD)

API Development: Knowledge of building and consuming APIs to integrate AI models into existing systems.

Version Control (Git): Proficiency in Git for collaborative development and managing codebases.

Candidates should already have knowledge to complete the following:-

25 Implementation completion checklist

Complete each item with status, owner, evidence link and acceptance date in the project tracker. Empty status means incomplete. This checklist is the release record, not a substitute for the detailed gate tests.

Gate

Required completion item

G0

Hardware, demand, source owners, tier model and licensing approved

G0

Pilot scope and held-out evaluation set approved

G1

Mac baseline, FileVault escrow, reboot and endpoint protection verified

G1

VLAN flows, deny rules, DNS, CA and identity/MFA verified

G2

Pinned native runtime and verified model pass smoke tests

G2

Separate tier data instances, keys and private bindings verified

G2

PostgreSQL, Qdrant and object restore smoke tests completed

G3

Source ACL inheritance, revocation and deletion completed

G3

Chunking/hybrid retrieval, exact identifiers and citations tested

G4

Gateway and broker deny by default; schemas and approval expiry tested

G4

n8n keys, credentials, retries and duplicate-event tests passed

G5

Functional, performance, safety and red-team criteria passed

G5

Full restore drill meets approved RPO and RTO

G5

Pilot owner accepts usefulness and user training

G6

Change window, backup point and rollback manifest approved

G6

Security, data owner and IT Head approve production cohort

G7

Runbooks, on call, asset register and ownership accepted

G7

Access review, patch cadence and capacity review scheduled

Natural Language Processing

• Java Programming Language

• Graph Query Language

• TypeScript Programming Language

• RESTful API

• Python Programming Language

• C# Programming Language

• Go Programming Language

• Rust Programming Language

• Django Software

• FastAPI Framework

• Flask Framework

• Machine Learning

• TensorFlow Software

• Transformers

• Scikit-Learn

• JavaScript Programming Language

• Angular

• Express.js

• Java Spring Boot

• Software Development

• ASP.NET

• ASP.NET Core

• Application Programming Interface (API)

• PHP Programming Language

• Data

• OAuth

• Microservices

• Computing

• Apache Kafka

• Health Insurance

• Information Technology

• Collaboration

• General Data Protection Regulation (GDPR)

• Auditing

• ISO Standards

• Access Control

• Generative Artificial Intelligence

• ChatGPT,

• Prompt Engineering

• Optimisation

• React Redux

• RxJS

• Cascading Style Sheets (CSS)

• Material UI

• Webpack

• Relational Database

• Microsoft Azure

• Oracle Applications

• PostgreSQL

• MySQL

• MongoDB

• Redis

• Apache Cassandra

• Amazon DynamoDB

• Elasticsearch Software

• Agentic RAG

  Candidates should understand the deployment of the following
  Phase 0 — security baseline

  Before deploying LLMs:

  • network VLANs;
  • FileVault;
  • MFA/SSO;
  • admin accounts;
  • firewall;
  • VPN;
  • backup;
  • secrets vault.

  Phase 1 — local inference

  Deploy:

  • llama.cpp/MLX;
  • models;
  • benchmark latency;
  • no external users yet.

  Phase 2 — secure RAG

  Add:

  • Qdrant;
  • Postgres;
  • document scanning;
  • access filtering;
  • provenance.

  Phase 3 — agents

  Add read-only agents.

  No actions.

  Phase 4 — controlled action

  Allow:

  • drafts;
  • CRM proposals;
  • task proposals.

  Human approval required.

  Phase 5 — limited automation

  Automate only low-risk deterministic actions.

  Phase 6 — monitoring / red-team

  Test:

  • prompt injection;
  • cross-tier retrieval;
  • malicious attachments;
  • API abuse;
  • agent loop;
  • credential leak;
  • data exfiltration.