<?xml version="1.0" encoding="UTF-8"?><oembed><type>video</type><version>1.0</version><html>&lt;iframe src=&quot;https://www.loom.com/embed/e8846453c7c74e6b8162478eb37b9713&quot; frameborder=&quot;0&quot; width=&quot;1280&quot; height=&quot;960&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>960</height><width>1280</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>960</thumbnail_height><thumbnail_width>1280</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/e8846453c7c74e6b8162478eb37b9713-6de67974e8bdbfb5.gif</thumbnail_url><duration>595.606</duration><title>BuildFlow AI — Construction Bid &amp;amp; Billing Workflow Assistant | Gen AI Portfolio | Apollo Sebastian</title><description>🏗️ BUILDFLOW AI
Construction Bid &amp; Billing Workflow Assistant
Powered by AWS Bedrock (Claude Haiku)  ·  Generative AI Bootcamp  ·  Instructor: Matt Dancho
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🚨 THE PROBLEM
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Small to mid-size general contractors lose 10–30% of potential revenue every year from broken back-office workflows. Bid preparation alone takes
8–15 hours per project — that is $600 to $2,250 in estimator labor before winning a single contract. No standard Scope of Work. No risk analysis.
No automated billing.

BuildFlow AI solves all of it — from one project description — in under 15 minutes.

  ⏱  8–15 hours per bid → under 15 minutes
  💸  Up to $2,250 in labor saved per project
  📉  10–30% revenue loss recovered annually

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

📄 WHAT IT GENERATES — 4 DOCUMENTS + AI CHAT (ONE PROJECT DESCRIPTION)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

TAB 1  ·  Trade-Level Bid Estimate
         Itemized by trade — Category, Description, Unit, Qty, Unit Cost, Total Subtotal  ·  Contingency %  ·  Total Bid card  ·  Download JSON

TAB 2  ·  Scope of Work (7-Section Legally Protective Document)
         Project Overview  ·  Inclusions  ·  Exclusions  ·  Milestones
         Payment Terms  ·  Change Order Policy  ·  Warranties  ·  Download TXT

TAB 3  ·  AIA-Style Progress Invoice
         Completion % slider (10–100%)  ·  Previously Billed input 4 live metric cards  ·  Automatic 10% retainage  ·  Download JSON

TAB 4  ·  Risk &amp; Change Order Report
         Risk registry: Category  ·  Severity  ·  Probability  ·  Mitigation
         Change Order table with cost impacts  ·  Contractor Protections  ·  Download JSON

TAB 5  ·  GC AI Assistant — Live Multi-Turn Chat  ⭐
         Full project context injected: client, location, sqft, timeline, total bid 16 sample questions across 4 categories
         Powered by AWS Bedrock Converse API

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

⚙️  AI TECHNIQUE — STRUCTURED PROMPT CHAINING + JSON EXTRACTION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Clinic 2 — Generative AI Bootcamp with Matt Dancho

3-step sequential chain — each step passes its output as context to the next:

  generate_bid()  →  generate_scope()  →  generate_risk_flags()

Animated progress bar shows all three AWS Bedrock calls running live. Claude Haiku returns structured JSON extracted with _strip_fences() and
_safe_json_loads() with built-in repair logic for clean rendering.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🎯  COURSE SKILLS DEMONSTRATED — BOOTCAMP RUBRIC
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

C1 · Structured Prompt Chaining                                    10 pts
     3-step chain: generate_bid → generate_scope → generate_risk_flags. Each step&apos;s JSON output feeds the next as live context.
     Animated progress bar confirms all 3 Bedrock calls in real time.

C2 · JSON Extraction                                               10 pts
     Claude Haiku returns structured JSON inside markdown code fences. App uses _strip_fences() + _safe_json_loads() with repair logic.
     Bid Estimate table and Risk Registry rendered entirely from parsed JSON.

C3 · AI Chat Completions — AWS Bedrock Converse API                 8 pts
     Tab 5 GC AI Assistant: run_assistant() → _bedrock_converse(). Full project context (client · location · sqft · timeline · total bid)
     + full conversation history injected into system prompt. Multi-turn.

C4 · Cloud Deployment — Live on Render.com                          6 pts
     Journey: AWS App Runner (shut down Apr 30, 2026 — external event) → Azure App Service (CI/CD built; WebSocket/Gunicorn incompatibility)
            → Render.com — deployed first attempt, GitHub auto-deploy active. 
     Live at: buildflow-ai-aws.onrender.com

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🔗  LIVE LINKS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

☁️  Live App (Render.com — Primary Cloud):
    https://buildflow-ai-aws.onrender.com

📺  Live App (Streamlit Mirror):
    https://buildflowai.streamlit.app

💻  GitHub:
    github.com/ASebastianAiX/buildflow-ai-aws

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Generative AI Bootcamp  |  Instructor: Matt Dancho  |  Apollo Sebastian  |  May 2026
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━</description></oembed>