Project Walkthrough: Building a Secure Banking AI Assistant
Build a secure banking AI assistant: policy RAG, entitled customer lookups, PII redaction, audit trail, red-teaming, evaluation and private deployment.
Learn
Guides, career roadmaps and explainers from Cloud Soft Solutions trainers. 667 articles.
Build a secure banking AI assistant: policy RAG, entitled customer lookups, PII redaction, audit trail, red-teaming, evaluation and private deployment.
Build a GitHub AI agent for first-pass PR reviews: webhooks, context assembly, MCP tools, comment-only App permissions, injection defence, noise control, evals.
Build a DevOps AI agent for on-call: alert intake, read-only MCP tools, runbook RAG, approved rollbacks, blast-radius controls and replay-based evaluation.
Build a multi-agent system project for employee onboarding: supervisor agent in LangGraph, message contracts, failure handling, approvals, tracing and evals.
AI agent frameworks comparison for enterprise teams: LangGraph vs CrewAI, OpenAI Agents SDK, Microsoft Agent Framework, Google ADK, Bedrock and no framework.
Build a SOC AI assistant that enriches alerts, maps MITRE ATT&CK, drafts case notes and proposes containment for approval, with tenant isolation and evaluation.
A document intelligence project end to end: OCR, LLM extraction to JSON, validation, confidence routing, human review, evaluation and cost per document.
Build a finance AI agent for accounts payable: invoice extraction, three-way match with finance-set tolerances, duplicate and fraud controls, approvals, evals.
Context engineering explained: what goes into an LLM's context window, system prompt design, few-shot, ordering, budget, memory, evals and security pitfalls.
What are tokens in an LLM? A beginner guide to subword tokenization, tokens vs words, context windows, token limits, cached tokens and how LLM pricing works.
Build an HR AI agent end to end: personalised policy RAG, own-data HRIS tools, confirmed leave requests, sensitive-topic routing, privacy, evaluation and ROI.
AI agent memory explained: working memory, checkpoints, long-term user memory, episodic memory, write and read policies, privacy, poisoning risks and testing.
Build a ServiceNow AI agent with an MCP server, LangGraph approval interrupts, per-user OAuth scoping, triage evaluation, deployment and an ROI method.
What is LLMOps? The LLMOps lifecycle, what gets versioned, LLMOps vs MLOps vs DevOps vs AIOps, tools by stage, team roles and maturity levels, explained.
A2A protocol explained: how the Agent2Agent standard lets AI agents discover each other and share tasks, how A2A differs from MCP, enterprise use and security.
Why LLM hallucinations happen, the main types with examples, why RAG does not eliminate them, and the engineering controls teams use to reduce them.
Azure for AI engineers: Azure OpenAI deployments, AI Foundry, AI Search RAG, Entra ID, Private Endpoints and Key Vault, with a private RAG reference deployment.
Build an AI customer support agent end to end: identity verification, RAG, order and return tools via MCP, human handoff, multilingual support, evals and ROI.
Practitioner guide to LLM cost optimization: where AI spend comes from, a cost-per-task model, routing, caching, batch, agent limits and GenAI FinOps.
Agentic RAG explained: query planning, routing across vector, SQL and API sources, corrective self-checks, stopping rules, costs, permissions and evaluation.
FDE vs prompt engineer compared: scope, code depth, ownership and careers, what prompting can and can't fix, and a concrete path from prompt skills to FDE.
Enterprise AI governance that enables shipping: risk tiers, roles, lifecycle gates, documentation, human oversight, incident response and a starter checklist.
Docker for AI applications: a FastAPI LLM Dockerfile, image size traps with CUDA and model weights, GPU containers, docker compose for RAG, secrets and SBOMs.
Terraform for AI infrastructure: private endpoints, scoped IAM, vector stores, Bedrock and Azure OpenAI, remote state, policy-as-code, drift and OIDC CI/CD.