Skip to content

Commit 90329d5

Browse files
gobinathmgobinathm
committed
Publishing other domain names for AIP-C01
1 parent 2032c77 commit 90329d5

9 files changed

Lines changed: 1212 additions & 99 deletions

File tree

‎.vitepress/config.mts‎

Lines changed: 23 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -140,7 +140,8 @@ export default defineConfig({
140140
{
141141
text: 'AWS',
142142
items: [
143-
{ text: 'MLA-C01: ML Engineer', link: '/certifications/aws/mla-c01/' },
143+
{ text: "AIP-C01: Generative AI Developer – Professional", link: "/certifications/aws/aip-c01/" },
144+
{ text: 'MLA-C01: Machine Learning Engineer', link: '/certifications/aws/mla-c01/' },
144145
{ text: 'AIF-C01: AI Practitioner', link: '/certifications/aws/aif-c01/' },
145146
{ text: 'CLF-C02: Cloud Practitioner', link: '/certifications/aws/clf-c02/' }
146147
]
@@ -312,8 +313,29 @@ export default defineConfig({
312313
items: [
313314
{ text: 'Overview', link: '/certifications/aws/mla-c01/' },
314315
{ text: 'Exam Objectives', link: '/certifications/aws/mla-c01/objectives' },
316+
{ text: 'Study Notes', link: '/certifications/aws/mla-c01/notes' },
317+
{ text: 'Quick Refresher', link: '/certifications/aws/mla-c01/quick-refresher' },
318+
{ text: 'Exam Tips', link: '/certifications/aws/mla-c01/exam-tips' },
315319
{ text: 'Resources', link: '/certifications/aws/mla-c01/resources' }
316320
]
321+
},
322+
{
323+
text: 'AIP-C01: Generative AI Developer – Professional',
324+
collapsed: true,
325+
items: [
326+
{ text: 'Overview', link: '/certifications/aws/aip-c01/' },
327+
{ text: 'Domain 1: FM Integration', link: '/certifications/aws/aip-c01/domain-1' },
328+
{ text: 'Domain 2: Implementation', link: '/certifications/aws/aip-c01/domain-2' },
329+
{ text: 'Domain 3: AI Safety & Governance', link: '/certifications/aws/aip-c01/domain-3' },
330+
{ text: 'Domain 4: Operational Efficiency', link: '/certifications/aws/aip-c01/domain-4' },
331+
{ text: 'Domain 5: Testing & Validation', link: '/certifications/aws/aip-c01/domain-5' },
332+
{ text: 'Exam Objectives', link: '/certifications/aws/aip-c01/objectives' },
333+
{ text: 'Study Notes', link: '/certifications/aws/aip-c01/notes' },
334+
{ text: 'Exam Guide', link: '/certifications/aws/aip-c01/exam-guide' },
335+
{ text: 'Cheatsheet', link: '/certifications/aws/aip-c01/cheatsheet' },
336+
{ text: 'Quick Refresher', link: '/certifications/aws/aip-c01/quick-refresher' },
337+
{ text: 'Exam Tips', link: '/certifications/aws/aip-c01/exam-tips' }
338+
]
317339
}
318340
]
319341
},

‎certifications/aws/aip-c01/cheatsheet.md‎

Lines changed: 152 additions & 17 deletions
Original file line numberDiff line numberDiff line change
@@ -1,32 +1,167 @@
11
---
22
title: "AIP-C01: Cheatsheet"
3-
description: "Quick reference and decision rules for the AWS Certified Generative AI Developer – Professional exam"
3+
description: "One-page exam day reference for AIP-C01 AWS Certified Generative AI Developer – Professional"
44
head:
55
- - meta
66
- name: keywords
7-
content: aip-c01, aws, bedrock, cheatsheet, quick reference, invokemodel, guardrails, rag, provisioned throughput, opensearch, vector store
7+
content: aip-c01, aws, bedrock, cheatsheet, quick reference, invokemodel, guardrails, rag, provisioned throughput, opensearch, vector store, exam day
88
---
99

10-
# AIP-C01 Cheatsheet
10+
# AIP-C01: Cheatsheet
1111

12-
[← Back to Overview](./index.md)
12+
[← Overview](./index.md) · [← Exam Guide](./exam-guide.md)
1313

14-
::: tip Quick Reference
15-
Decision rules and service comparisons for last-minute review before the AIP-C01 exam.
14+
::: danger Exam Day Reference
15+
Review this page 5 minutes before the exam.
1616
:::
1717

18-
*Content coming soon as I progress through my studies.*
18+
---
19+
20+
## Foundation Model Quick Reference
21+
22+
| FM | Vendor | Key Strength | Best For |
23+
|---|---|---|---|
24+
| **Claude** | Anthropic | 200k token context, reasoning | Long docs, complex reasoning |
25+
| **Llama** | Meta | Open-source, fine-tunable | Custom fine-tuning |
26+
| **Mistral** | Mistral AI | Efficient, fast | Cost-efficient inference |
27+
| **Titan** | AWS | AWS-native, embeddings | RAG embeddings, summarization |
28+
| **Cohere Embed** | Cohere | Multilingual embeddings | Multilingual RAG |
29+
30+
---
31+
32+
## Bedrock API Comparison
33+
34+
| API | Delivery | Use Case |
35+
|---|---|---|
36+
| `InvokeModel` | Synchronous (full response) | Simple query-response |
37+
| `InvokeModelWithResponseStream` | Streaming (token by token) | Low-latency UX / chat |
38+
| `InvokeAgent` | Streaming + trace | Multi-step agentic workflows |
39+
40+
---
41+
42+
## Agents vs. Knowledge Bases
43+
44+
| | Knowledge Base (RAG only) | Bedrock Agents |
45+
|---|---|---|
46+
| **External API calls** | No | Yes (via Action Groups + Lambda) |
47+
| **Multi-step reasoning** | No | Yes |
48+
| **Document retrieval** | Yes | Yes (optional) |
49+
| **Best for** | Static knowledge Q&A | Dynamic workflows, tool use |
50+
51+
---
52+
53+
## Guardrails — Four Filter Types
54+
55+
| Filter Type | What It Controls |
56+
|---|---|
57+
| **Content Filters** | Harmful categories: hate, violence, sexual content, insults |
58+
| **Denied Topics** | Topics the model must refuse to discuss |
59+
| **Word Filters** | Exact word/phrase blocklists |
60+
| **PII Redaction** | Names, emails, SSNs, credit cards — Redact or Block |
61+
62+
**Key rules:**
63+
- Guardrails apply to **both inputs AND outputs**
64+
- Must be explicitly applied per API call via `guardrailIdentifier` + `guardrailVersion`
65+
- PII modes: **Redact** (mask with placeholder) vs. **Block** (reject request/response)
66+
67+
---
68+
69+
## Vector Store Options
70+
71+
| Store | Type | Use When |
72+
|---|---|---|
73+
| **Amazon OpenSearch Serverless** | Managed, serverless | Bedrock Knowledge Bases (default) |
74+
| **Aurora PostgreSQL (pgvector)** | RDS extension | Existing PostgreSQL infrastructure |
75+
| **Amazon Kendra** | Enterprise search | NLP-powered enterprise retrieval |
76+
77+
---
78+
79+
## Chunking Strategies
80+
81+
| Strategy | Best For | Trade-off |
82+
|---|---|---|
83+
| **Fixed-size** | Uniform documents | May break context at boundaries |
84+
| **Fixed-size + overlap** | Preserving cross-boundary context | Higher storage cost |
85+
| **Semantic** | Varied, long-form content | Higher processing complexity |
86+
| **Hierarchical** | Complex docs: broad + fine retrieval | More complex retrieval logic |
87+
88+
---
89+
90+
## Provisioned Throughput vs. On-Demand
91+
92+
| | Provisioned Throughput (PTU) | On-Demand |
93+
|---|---|---|
94+
| **Traffic** | Predictable, 24/7 | Sporadic, variable |
95+
| **Pricing** | Fixed (per MU/hour) | Per token |
96+
| **Commitment** | 1 month or 6 months | None |
97+
| **Best for** | Production steady-state | Dev/test, bursts |
98+
99+
Batch Inference = ~50% cheaper than on-demand for non-real-time high-volume jobs.
19100

20101
---
21102

22-
## Amazon Bedrock APIs
23-
- `InvokeModel`: Synchronous call to get a complete response.
24-
- `InvokeModelWithResponseStream`: Asynchronous/streaming response for low-latency feel.
103+
## Model Evaluation Metrics
104+
105+
| Metric | What It Measures |
106+
|---|---|
107+
| **Groundedness** | Response supported by retrieved context? (detects hallucinations) |
108+
| **Relevance** | Response answers the user's question? |
109+
| **Accuracy** | Factually correct? |
110+
| **Fluency** | Well-written and natural? |
111+
112+
---
113+
114+
## Quick Decision Rules
115+
116+
**Vector store for Knowledge Base?**
117+
→ Amazon OpenSearch Serverless (default) · pgvector on Aurora (alternative)
118+
119+
**Multi-step reasoning + tool use?**
120+
→ Bedrock Agents + Action Groups (Lambda)
121+
122+
**Content moderation / PII?**
123+
→ Guardrails for Amazon Bedrock
124+
125+
**Audit trail for compliance?**
126+
→ AWS CloudTrail (not CloudWatch)
127+
128+
**Operational monitoring (latency, errors, token counts)?**
129+
→ Amazon CloudWatch
130+
131+
**Predictable 24/7 throughput?**
132+
→ Provisioned Throughput (PTU) — 1 or 6 month commitment
133+
134+
**Bulk, non-real-time inference?**
135+
→ Batch Inference
136+
137+
**Private connectivity to Bedrock API?**
138+
→ VPC Endpoint — `bedrock-runtime` for inference calls
139+
140+
**Knowledge changes frequently / need traceability?**
141+
→ RAG (Knowledge Bases), not fine-tuning
142+
143+
**Detect hallucinations in RAG?**
144+
→ Groundedness metric in Model Evaluation
145+
146+
**Log all prompts and responses for AI governance?**
147+
→ Model Invocation Logging (to S3 or CloudWatch Logs)
148+
149+
---
150+
151+
## Key Terminology
152+
153+
- **FM**: Foundation Model — pre-trained large AI model (Claude, Llama, Titan, etc.)
154+
- **RAG**: Retrieval-Augmented Generation — FM inference + vector store retrieval
155+
- **PTU**: Provisioned Throughput Unit — reserved Bedrock model capacity
156+
- **MU**: Model Unit — unit of PTU capacity purchased
157+
- **PII**: Personally Identifiable Information — data that identifies an individual
158+
- **Groundedness**: Metric measuring how well a response is grounded in retrieved context
159+
- **Hallucination**: FM generating information not present in the provided context
160+
- **Action Group**: Lambda function exposed to a Bedrock Agent for tool use
161+
- **OpenSearch Serverless**: Managed, serverless vector store used by Bedrock Knowledge Bases
162+
- **Batch Inference**: Asynchronous bulk FM inference via S3 JSONL input/output
163+
- **Model Invocation Logging**: Bedrock feature that logs all prompts + responses to S3/CloudWatch Logs
164+
165+
---
25166

26-
## Decision Rules
27-
| Requirement | Service/Feature |
28-
|-------------|-----------------|
29-
| Predictable Throughput | Provisioned Throughput |
30-
| Data Residency | Private Endpoints + Region Selection |
31-
| Content Moderation | Guardrails for Amazon Bedrock |
32-
| Knowledge Base | Amazon OpenSearch Serverless |
167+
[← Overview](./index.md) · [← Exam Guide](./exam-guide.md)

0 commit comments

Comments
 (0)