Official final judge pack

AI E-Commerce Chat Interface Prototype Challenge

Professional judge pack with SWOT, bug/logic review, risk matrix, maturity rating, live presentation script, candidate feedback, and official panel score alignment.

Scoring scope: uploaded source/docs/slides plus panel observations. Runtime status is treated as not verified unless shown live, not automatic failure.

Official score summary

This section uses the updated panel score sheet as the official final score. My source/code review is used as supporting evidence and improvement feedback.

Rank	Candidate	Product	Final score	Official judge direction	Key reason
1	🥇 Alex	ShopMate AI	92.33	Champion	Best overall balance: clear explanation, strong core structure, strong AI chatbot flow, and convincing privacy demonstration.
2	🥈 Raj	ShopEase / TARA AI	87.33	First Runner-Up	Strong code, strong product/design thinking, and useful AI extension concept. Main deduction is privacy/session behavior and missing formal UML/structure docs.
3	🥉 Tay	ShopAI Agent	78.00	Second Runner-Up	Strongest submitted artifact and architecture depth, but official score drops because live/presentation performance and some AI functions were not reliable enough.
4	④ Mekky	AI Chat Interface	75.00	Fourth Place	Good improvement and good voice capability, but the chatbot response UI was text-heavy and the demo data was not sufficient to prove full functions.
5	⑤ Phonesai	POS + AI Chatbot	66.33	Fifth Place	Practical POS/chatbot build with some privacy shown, but UI, AI accuracy, and security/ownership concerns keep the score moderate.
6	⑥ Da	Ecommerce + Local AI	56.67	Sixth Place	Technically interesting local AI/server direction, but many basic ecommerce challenge features and product-pitch elements are still weak compared with others.

Important score adjustment: Tay still has one of the strongest artifact packages, but the official final score is lower because panel/live performance and specific AI functions did not demonstrate reliably enough. Alex and Raj remain the top group because the challenge also rewards clear explanation, ownership, and practical product understanding.

Winner announcement board

Use this for the final internal announcement or presentation closing slide.

🥈

Silver - First Runner-Up

Raj

ShopEase / TARA AI

87.33

Strong code, strong product/design thinking, and useful AI extension concept. Main deduction is privacy/session behavior and missing formal UML/structure docs.

🥇

Gold - Champion

Alex

ShopMate AI

92.33

Best overall balance: clear explanation, strong core structure, strong AI chatbot flow, and convincing privacy demonstration.

🥉

Bronze - Second Runner-Up

Tay

ShopAI Agent

78.00

Strongest submitted artifact and architecture depth, but official score drops because live/presentation performance and some AI functions were not reliable enough.

Announcement wording: Congratulations to Alex for 1st place with ShopMate AI, Raj for 2nd place with ShopEase / TARA AI, and Tay for 3rd place with ShopAI Agent. The winning group showed the strongest combination of product thinking, AI-assisted development, data/privacy awareness, and presentation ownership.

Visual preview / runtime attempt

These previews are added so non-technical judges can see what was submitted. They are not full runtime proof unless the candidate demonstrated live.

🥇 Alex - ShopMate AI

Preview from submitted final presentation PDF.

🥈 Raj - ShopEase / TARA AI

Raj frontend started locally; preview captured from a static SSR snapshot. Full runtime remains not verified.

🥉 Tay - ShopAI Agent

Preview from submitted ShopAI Agent presentation.

④ Mekky - AI Chat Interface

Preview from submitted ERD/diagram artifact.

⑤ Phonesai - POS + AI Chatbot

Preview from submitted final presentation PPTX.

⑥ Da - Ecommerce + Local AI

Preview generated from submitted source/database structure because no rich final UI preview was available.

Runtime note: Raj frontend could be started locally enough to capture a static SSR preview. The other candidate previews come from submitted slides, diagrams, or source structure because full runtime required missing dependencies, backend services, databases, or live credentials. Therefore, runtime remains not verified unless shown in the final presentation.

Final criteria used in the panel score sheet

30%

Technical Functionality

Prototype works and demonstrates ecommerce functions such as stock checking, product inquiry, shipment/order response.

20%

AI Prompt Design & Response Quality

Prompts and AI responses are accurate, useful, business-oriented, and grounded in real/dummy data.

15%

Database & Data Management

Structured data can manage products, stock, orders, shipments and user privacy.

15%

User-Friendliness

Interface is clear and suitable for ecommerce users and panel demo.

10%

Product Concept & Practicality

Product idea has business value and realistic application.

10%

Presentation & Progress Management

Weekly progress, final explanation, confidence, and Q&A.

Source review vs panel/live review

Candidate	Source/docs review	Panel/live observation impact	Final interpretation
Alex	Strong and consistent with presentation.	Panel score confirms strong explanation and privacy demonstration.	Source and live performance align; clear winner.
Raj	Strong code and product/design direction; missing formal UML/ERD docs.	Panel saw strong features but also privacy/session issue and extension not fully working.	Top group, but privacy/session fix is mandatory.
Tay	Strongest architecture/document package.	Live score reduced due demo/presentation and AI function issues.	Good artifact, but official ranking is 3rd.
Mekky	Good progress docs and diagrams.	Panel praised improvement and Lao/English voice but noted text-heavy UI.	Good progress; needs stronger UI/data and code enforcement.
Phonesai	Practical POS/chatbot/source package.	Panel found weak UI and accuracy issues.	Basic pass with practical value; needs security and UX improvement.
Da	Backend/frontend/SQL exist but weekly evidence weak.	Panel found many basics still missing.	Technical curiosity but below expected challenge level.

Risk matrix

Candidate	Risk level	Main risk	Judge action
Alex	Medium	Earlier source review suggested order tracking/privacy should be confirmed at backend level, not only by UI/prompt.	Use candidate-specific Q&A and privacy test before final acceptance.
Raj	Medium	Panel observed that after User A signed out, chat history still displayed. This is a privacy/session leak for a chat-based commerce app.	Use candidate-specific Q&A and privacy test before final acceptance.
Tay	Medium	Panel observed refund flow did not demonstrate correctly. This affects one of the headline support use cases.	Use candidate-specific Q&A and privacy test before final acceptance.
Mekky	Medium-High	Earlier source review indicated conversation/message ownership checks may not be strict enough.	Use candidate-specific Q&A and privacy test before final acceptance.
Phonesai	Medium-High	Earlier source review indicated the chatbot could rely on client-provided user ID/header.	Use candidate-specific Q&A and privacy test before final acceptance.
Da	High	The project direction did not fully match ecommerce chat/product-stock-order use case.	Use candidate-specific Q&A and privacy test before final acceptance.

Rank 1 · ShopMate AI

🥇 Alex

Best overall balance: clear explanation, strong core structure, strong AI chatbot flow, and convincing privacy demonstration.

Level 4 - staging-ready directionRisk: MediumOfficial panel score

92.33

/100

Official score breakdown

Criteria	Weight	Week 4	Week 1-3	Final used
Technical Functionality	30	29	27	28
AI Prompt Design & Response Quality	20	20	17	18.5
Database & Data Management	15	15	13	14
User-Friendliness	15	13	14	13.5
Product Concept & Practicality	10	9	9	9
Presentation & Progress Management	10	10	9	9.33
Total	100			92.33

Challenge match check

Ecommerce chat interfaceEvidence seen Product / stock inquiryEvidence seen Order / invoice or tracking flowEvidence seen Database / structured dataEvidence seen AI prompt or agent logicEvidence seen User A / User B privacyCheck live Weekly progress evidenceEvidence seen Clear final presentationCheck live

Panel comments to carry into report

Basic required features were fulfilled with strong core structure.
AI prompt design was clearly presented, and chatbot functionality worked for user-platform and seller-platform scenarios.
User data privacy logic was strongly shown during presentation.
Improvement: AI innovation can be expanded beyond the strong core flow.

Evidence from submitted files

Final source contains separate Express backend and Nuxt frontend folders.
Database exports and technical documentation were included.
Week 1 and Week 2 presentation PDFs were included; Week 3 folder was empty in the uploaded archive.
Code/source structure supports the panel observation that the core was well organized.

SWOT analysis

Strengths

Strongest explanation and structure in the panel view.
Clear user/seller platform thinking.
Good UI direction and mobile-supported concept.
Good privacy demonstration compared with most candidates.

Weaknesses

Innovation could go further beyond required features.
Week 3 artifact was not visible in the uploaded folder.
Need a final formal privacy test script or automated test evidence.

Opportunities

Add advanced AI agent actions such as campaign suggestions, reorder prediction, or automated support triage.
Add richer recommendation rules and structured AI response cards.
Document the exact route/service where user-data isolation is enforced.

Threats

If the privacy flow is only demonstrated visually but not documented in code-level tests, future reuse may carry risk.
High score depends on continuing to show live ownership of the code during Q&A.

Bug / logic / security review

Severity	Issue	Why it matters	Recommended fix
Medium	Public/order privacy confirmation	Earlier source review suggested order tracking/privacy should be confirmed at backend level, not only by UI/prompt.	Add backend owner checks and a User A/User B privacy test.
Low	Innovation ceiling	The core is strong, but AI features can become predictable if only basic ecommerce questions are supported.	Add more agent-like workflows and richer product reasoning.

Final presentation notes / questions to ask

Show the backend/API route where User A is filtered from User B orders. Demo one user flow from product search to invoice/order inquiry. Explain what AI helped with and what you personally designed. Show how seller-side chatbot differs from customer-side chatbot.

Feedback to candidate

Excellent final result. Keep your clear structure and presentation style. To improve further, show more innovation in the AI layer and provide a simple code-level privacy test that proves User A cannot access User B data.

Rank 2 · ShopEase / TARA AI

🥈 Raj

Strong code, strong product/design thinking, and useful AI extension concept. Main deduction is privacy/session behavior and missing formal UML/structure docs.

Level 4 - staging-ready directionRisk: MediumOfficial panel score

87.33

/100

Official score breakdown

Criteria	Weight	Week 4	Week 1-3	Final used
Technical Functionality	30	28	27	27.5
AI Prompt Design & Response Quality	20	17	17	17
Database & Data Management	15	10	13	11.5
User-Friendliness	15	13	14	13.5
Product Concept & Practicality	10	10	9	9.5
Presentation & Progress Management	10	9	8	8.33
Total	100			87.33

Challenge match check

Ecommerce chat interfaceEvidence seen Product / stock inquiryEvidence seen Order / invoice or tracking flowEvidence seen Database / structured dataEvidence seen AI prompt or agent logicEvidence seen User A / User B privacyIssue found Weekly progress evidenceEvidence seen Clear final presentationCheck live

Panel comments to carry into report

Basic features were fulfilled together with many additional features, but the core flow was not as strong as Alex.
AI chatbot as extension across platforms is an interesting practical direction, but it did not fully work during the observed presentation.
Innovative ideas include security alert when a user attempts to access another user’s data and seller dashboard/campaign recommendations.
Panel issue: User A signed out but chat history still displayed, creating a privacy/session concern.

Evidence from submitted files

Source includes backend, frontend, extension, product assets, and data folders.
Backend has controllers, routes, services, models, middleware, validators, seed and config folders.
Models and services show customer/vendor/admin roles, products, cart, wishlist, orders, reviews, vendor dashboard and AI chat tools.
No clear UML/ERD/system-structure document was found in the Raj zip, so non-technical judges need live explanation.

SWOT analysis

Strengths

Strong source-code organization and broad product feature coverage.
Good design/product focus and clear code understanding in panel observation.
AI extension direction can be reused outside one ecommerce app.
Seller-side campaign recommendation is business-relevant.

Weaknesses

Formal UML/ERD/architecture docs are missing.
Session/privacy issue: chat history visible after sign-out must be fixed.
Extension did not fully work during observed presentation.
Submitted package included .env files; secrets should never be submitted if real.

Opportunities

Turn the extension concept into a clean B2B product differentiator.
Add simple architecture/ERD diagrams to support non-technical judging.
Strengthen user session clearing and privacy boundary tests.
Use seller dashboard AI as a real company prototype direction.

Threats

A visible privacy/session leak can reduce trust even if backend code is strong.
If .env values are real, they must be rotated and removed from future submissions.
If extension demo fails, judges may see broad features as unfinished.

Bug / logic / security review

Severity	Issue	Why it matters	Recommended fix
High	Chat history visible after sign-out	Panel observed that after User A signed out, chat history still displayed. This is a privacy/session leak for a chat-based commerce app.	Clear client state on logout; invalidate session; re-check auth before loading chat history.
Medium	Missing UML/ERD documentation	Technical structure exists in code, but non-technical judges cannot easily understand it without diagrams.	Add one system architecture diagram, one ERD, and one AI flow diagram.
Medium	.env files included	Environment files were included in the archive. If they contain real keys, this is a security hygiene issue.	Remove .env from submissions and rotate any real keys.

Final presentation notes / questions to ask

Show the logout flow and prove chat history is cleared after sign-out. Show exact code path that prevents User A from reading User B data. Explain how TARA extension connects to ecommerce platforms. Show the seller campaign recommendation feature and explain the data used.

Feedback to candidate

Very strong work and clearly top group. The biggest improvement is to make the privacy/session behavior bulletproof and document the structure visually. Your code shows strong ability, but the final pack should make it easier for non-technical judges to understand.

Rank 3 · ShopAI Agent

🥉 Tay

Strongest submitted artifact and architecture depth, but official score drops because live/presentation performance and some AI functions were not reliable enough.

Level 4 artifact / Level 3 live resultRisk: MediumOfficial panel score

78.00

/100

Official score breakdown

Criteria	Weight	Week 4	Week 1-3	Final used
Technical Functionality	30	18	27	22.5
AI Prompt Design & Response Quality	20	10	19	14.5
Database & Data Management	15	15	14	14.5
User-Friendliness	15	12	13	12.5
Product Concept & Practicality	10	7	9	8
Presentation & Progress Management	10	4	8	6
Total	100			78.00

Challenge match check

Panel comments to carry into report

Rich features were available, but only some basic features were fulfilled strongly during presentation.
Smart AI search and image search were not fully precise; example: AirTag query returned broad Apple products.
AI voice was available only in English.
Privacy was demonstrated when not logged in and AI responses based on changing data were accurate.
Refund prompt failed to demonstrate properly; presentation needed improvement.

Evidence from submitted files

Submitted package includes presentation, prompt document, full source folder and documentation.
Source structure includes backend/admin/vendor/mobile/shared/Prisma-style architecture, with rich AI tool and card-rendering concept.
Prompt rules emphasize real data, no invented prices/stock/orders, tool calls, confirmation before mutating actions, and cart/order rendering.
Runtime was not fully verified in sandbox because the package required services/dependencies; panel live result is therefore the official deciding factor.

SWOT analysis

Strengths

Best architecture/documentation artifact depth among the candidates.
Strong AI tool-use concept: model reads real rows and renders structured UI cards.
Good database/privacy thinking in the written design.
Broad ecommerce support coverage: recommendations, cart, refund, tracking, promotions.

Weaknesses

Live presentation score was weak compared with artifact quality.
Some AI functions were inaccurate or not demonstrated successfully.
Scope is broad, which can make the core ecommerce flow less convincing.
Prompt document has competing instructions around product card vs text response style.

Opportunities

Narrow the demo to one reliable core flow.
Improve image search precision and refund flow reliability.
Use the strong diagrams more clearly in presentation.
Convert broad feature list into staged roadmap.

Threats

A strong codebase that cannot be defended live may be judged as AI-generated without ownership.
Broad feature claims can backfire if one or two key demos fail.
If prompt rules conflict, chatbot behavior may be inconsistent.

Bug / logic / security review

Severity	Issue	Why it matters	Recommended fix
High	Refund prompt/demo failed	Panel observed refund flow did not demonstrate correctly. This affects one of the headline support use cases.	Simplify refund prompt and tool path; demo with one seeded eligible order.
Medium	Image search precision	AirTag search returned broad Apple products, reducing trust in AI search quality.	Add category/entity filtering and result confidence threshold.
Medium	Presentation ownership gap	Artifact quality is high but presentation score is low. Judges need proof the candidate understands the code.	Use a shorter demo script and be ready to explain exact source files.

Final presentation notes / questions to ask

Show the exact product search tool and how it prevents fake product/price answers. Demo refund flow using one seeded delivered order. Explain why image search returned broad Apple products and how you would fix it. Show the database table/schema that links user, order, product and chat.

Feedback to candidate

Your artifact is very strong, but final judging depends on what you can show and explain live. Reduce scope, make the main ecommerce flow reliable, fix refund/image search, and clearly defend how the AI tool layer works.

Rank 4 · AI Chat Interface

④ Mekky

Good improvement and good voice capability, but the chatbot response UI was text-heavy and the demo data was not sufficient to prove full functions.

Level 3 - working prototype directionRisk: Medium-HighOfficial panel score

75.00

/100

Official score breakdown

Criteria	Weight	Week 4	Week 1-3	Final used
Technical Functionality	30	20	23	21.5
AI Prompt Design & Response Quality	20	13	15	14
Database & Data Management	15	15	10	12.5
User-Friendliness	15	10	12	11
Product Concept & Practicality	10	8	8	8
Presentation & Progress Management	10	9	10	8
Total	100			75.00

Challenge match check

Panel comments to carry into report

Basic features were mostly fulfilled.
AI voice was highly functional in both Lao and English.
Prototype improved significantly compared with previous weeks.
AI responses were text-heavy and not arranged in a user-friendly way.
More dummy data and innovative AI response UI are needed.

Evidence from submitted files

Week 1 source and presentation files were included.
Week 2 includes business flow, privacy flow, ERD, AI thinking and real AI flow diagrams.
Week 3 package includes backend/frontend and docs, showing visible progress.
Final package includes a presentation HTML and project files.

SWOT analysis

Strengths

Best visible week-by-week improvement evidence after Alex/Raj.
Strong diagram package for Week 2.
Voice AI in Lao and English is a real positive feature.
Understands privacy at documentation level.

Weaknesses

AI response display is too text-heavy and not commerce-card friendly.
Data volume not enough to demonstrate full chatbot functions.
Privacy enforcement in source needs to fully match the diagrams.
Pitch could be more product-focused.

Opportunities

Turn text responses into product/order/shipment cards.
Add more dummy orders/products/shipments to prove retrieval.
Use diagrams as a presentation strength.
Add backend ownership guards for chat/order routes.

Threats

If the system only explains privacy but does not enforce it in code, it can fail the core condition.
Text-heavy UI makes the product feel less ready than its logic.

Bug / logic / security review

Severity	Issue	Why it matters	Recommended fix
High	Privacy enforcement gap	Earlier source review indicated conversation/message ownership checks may not be strict enough.	Add authenticated ownership checks on chat/message routes.
Medium	Text-heavy AI output	Panel noted AI responses were not user-friendly.	Render product/order information as cards or structured blocks.
Medium	Insufficient demo data	Limited data prevents full proof of stock/order/shipment functions.	Seed more products, orders, shipments, and user-specific histories.

Final presentation notes / questions to ask

Show the Lao/English voice AI demo. Show the ERD/privacy diagram and then show the matching code. Show an AI response using structured product/order cards, not plain text. Explain what improved from Week 1 to Week 4.

Feedback to candidate

Good progress and strong effort. Your biggest improvement is not only adding more code, but making the chatbot answer visually and making privacy rules enforceable in the backend.

Rank 5 · POS + AI Chatbot

⑤ Phonesai

Practical POS/chatbot build with some privacy shown, but UI, AI accuracy, and security/ownership concerns keep the score moderate.

Level 3 - practical prototype with security riskRisk: Medium-HighOfficial panel score

66.33

/100

Official score breakdown

Criteria	Weight	Week 4	Week 1-3	Final used
Technical Functionality	30	15	24	19.5
AI Prompt Design & Response Quality	20	13	14	13.5
Database & Data Management	15	15	8	11.5
User-Friendliness	15	5	11	8
Product Concept & Practicality	10	7	8	7.5
Presentation & Progress Management	10	8	4	6.33
Total	100			66.33

Challenge match check

Ecommerce chat interfaceEvidence seen Product / stock inquiryEvidence seen Order / invoice or tracking flowEvidence seen Database / structured dataEvidence seen AI prompt or agent logicEvidence seen User A / User B privacyRisky Weekly progress evidenceEvidence seen Clear final presentationCheck live

Panel comments to carry into report

Basic features were partially fulfilled.
AI chatbot satisfied minimum requirements, and restricted data privacy was shown.
AI response was not fully accurate; example: asking for car displayed laptop.
UI was not complete; mostly the chatbot was properly done.
Data was not sufficient to demonstrate full chatbot functions.

Evidence from submitted files

Submitted final package includes chatbot-api, main-pos-api, pos-frontend, SQL database scripts, plan, and final presentation.
Chatbot logic includes product/stock/order direction.
Database scripts indicate real structure effort.
Earlier source review found user identity and secret-handling issues.

SWOT analysis

Strengths

Practical business direction close to POS/ecommerce operations.
Backend/API/database scripts show actual implementation effort.
Minimum chatbot and privacy demonstration were present.
Good starting point for internal POS assistant.

Weaknesses

UI score is low because only the chatbot was reasonably complete.
AI product matching accuracy is weak.
User identity model needs stronger authentication/ownership enforcement.
Some sensitive/private key material was present in submitted source review.

Opportunities

Improve AI product matching with controlled categories and real data lookup.
Build product/order/invoice UI around the chatbot.
Replace client-provided user identity with signed-in session/token.
Add more sample data to demonstrate chatbot capabilities.

Threats

If user ID can be faked from the client, privacy can fail.
Bad product matching reduces trust in AI answers.
Included keys/secrets can create security risk if reused.

Bug / logic / security review

Severity	Issue	Why it matters	Recommended fix
High	User identity trust issue	Earlier source review indicated the chatbot could rely on client-provided user ID/header.	Use authenticated session/token only; never trust browser-sent user ID.
High	Secrets/key hygiene	Private key material was included in the submitted archive during review.	Remove from repository and rotate if real.
Medium	AI answer mismatch	Panel saw irrelevant product matching such as car question returning laptop.	Add product category/entity matching and fallback when no product exists.

Final presentation notes / questions to ask

Show where logged-in user identity comes from. Show order/cancel/check-stock flow and prove it filters by owner. Ask the AI for a missing product and show it does not invent. Show product cards/order table instead of text-heavy answer.

Feedback to candidate

Good hands-on build, but the system needs stronger UI and safer ownership logic. Focus on reliable product matching, clear structured responses, and authentication before adding more features.

Rank 6 · Ecommerce + Local AI

⑥ Da

Technically interesting local AI/server direction, but many basic ecommerce challenge features and product-pitch elements are still weak compared with others.

Level 2-3 - technical explorationRisk: HighOfficial panel score

56.67

/100

Official score breakdown

Criteria	Weight	Week 4	Week 1-3	Final used
Technical Functionality	30	10	18	14
AI Prompt Design & Response Quality	20	12	12	12
Database & Data Management	15	15	9	12
User-Friendliness	15	5	8	6.5
Product Concept & Practicality	10	6	7	6.5
Presentation & Progress Management	10	7	7	5.67
Total	100			56.67

Challenge match check

Panel comments to carry into report

Many basic features still need to be fulfilled compared with other candidates.
AI chatbot satisfies minimum requirements, but responses are text-heavy.
Restricted privacy was shown, and database/data score improved in Week 4.
UI/product concept is simple and outdated compared with other submissions.
Extensive improvement is needed in UI and innovative AI features.

Evidence from submitted files

Submitted final package includes backend zip, frontend zip, SQL database and Week 1 repository document.
Week 2 and Week 3 folders were empty in the uploaded archive.
Backend/frontend folders show implementation effort, but product story is weak.
Local AI/server exploration was a plus, but not enough for the ecommerce chat product goal.

SWOT analysis

Strengths

Local AI/server exploration shows technical curiosity.
Database diagram/SQL is a positive point.
Some restricted privacy demonstration was present.
Backend/frontend source exists.

Weaknesses

Weakest weekly-progress evidence in uploaded files.
Product concept and UI are behind the others.
Focus was too broad or infrastructure-oriented instead of ecommerce chat.
Presentation was not non-technical friendly.

Opportunities

Rebuild around one clear flow: product -> stock -> order -> invoice -> chat.
Use local AI only as supporting feature, not the product itself.
Add simple visual UML/ERD and user flow diagrams.
Create richer demo data and structured response UI.

Threats

If the product remains infrastructure-focused, it will miss the challenge goal.
Weak weekly evidence suggests lower dedication/process compliance.
Too-simple UI may make the work look incomplete even if backend exists.

Bug / logic / security review

Severity	Issue	Why it matters	Recommended fix
High	Product alignment gap	The project direction did not fully match ecommerce chat/product-stock-order use case.	Focus all features on the required ecommerce user journey.
Medium	Text-heavy responses	Panel noted user-friendliness problems similar to lower candidates.	Use cards/tables for stock, order, invoice and shipment.
Medium	Weekly evidence missing	Week 2 and Week 3 folders were empty in the uploaded archive.	Maintain weekly progress documents and screenshots.

Final presentation notes / questions to ask

Show the database diagram and explain it in business language. Show one complete ecommerce flow with chatbot. Explain why local AI helps this product specifically. Show Week 2/Week 3 progress evidence if available.

Feedback to candidate

Good technical curiosity, especially around local AI, but the challenge is about a practical ecommerce chat prototype. Re-focus on product flow, visual explanation, and user-friendly demo before expanding infrastructure.

Live presentation script for all candidates

Use this exact flow to compare candidates fairly.

Required live flow

Open app and identify the product concept.
Login as User A.
Show product/stock list.
Ask chatbot: “What stock is available?”
Create order/invoice from selected product.
Ask chatbot: “Show my latest order/invoice.”
Ask shipment/order status.
Logout and login as User B.
Try to access User A order/chat/history.
System must reject or return no private data.
Show admin/seller view if included.
Explain what AI generated and what the developer verified.

Pass / fail gate

Privacy gate: User A/User B data leakage should cap database/privacy score.
AI grounding gate: AI must not invent products, prices, stock, orders or shipment status.
Ownership gate: Candidate must explain their code and prompt design.
Runtime gate: If not shown live, mark runtime as not verified, not automatic fail.

Core panel questions

AI understanding

What part did AI help you build?
What part did you personally design?
How do you prevent hallucinated product/order data?

Database & privacy

Where do you filter by user/customer/seller?
What happens if User B asks for User A order?
Is privacy enforced by prompt, backend, or both?

Product thinking

What real ecommerce problem does this solve?
How does it reduce support workload?
Which feature should become production first?

Winner announcement script

1st Place - Alex, ShopMate AI: strongest overall balance of core functionality, AI response quality, structure, presentation, and privacy demonstration.

2nd Place - Raj, ShopEase / TARA AI: strong product/design thinking, broad ecommerce functions, and innovative extension/seller-AI direction, with privacy/session fixes required.

3rd Place - Tay, ShopAI Agent: strongest architecture/documentation package and advanced AI-agent concept, with live demo and presentation reliability needing improvement.

Candidate feedback template

Topic	Suggested wording
Strength	Your strongest point is [specific feature/evidence]. This shows [mindset or capability].
Improvement	Your next improvement should be [specific fix], because it affects [business/user/privacy/demo quality].
AI usage	Continue using AI as an assistant, but make sure you can explain and verify every generated part.
Privacy	For any AI/database product, backend ownership checks are more important than prompt rules alone.