Uploaded February 2026 | Updated September 2026, 2 weeks ago
AI adoption isn't just about giving everyone a ChatGPT Plus account. It’s about infrastructure, security, and cost 🔓
In this talk, Andy Kuszyk, Staff Engineer at Typeform, shares how Typeform built a scalable AI ecosystem using AWS Bedrock and MCP servers. Learn how to democratize LLM access while keeping your spend under control and your engineering DNA AI-driven.
Instead of prescribing specific IDEs, learn how to build a "paved road" that empowers engineers to use their preferred tools—whether it's Cursor, Cline, or Emacs—while keeping security and billing centralized.
You will learn:
🔹 Architecture: Why Typeform chose AWS Bedrock over direct vendor integrations (Anthropic/OpenAI) to solve billing chaos.
🔹 Implementation: How to replace high-friction IAM/VPN requirements with self-serve Bedrock API Keys.
🔹 Federated Usage: How to technically support any AI tool via a unified token system without vendor lock-in.
🔹 Process Impact: How unlimited AI access transforms engineering hiring interviews and design-to-code prototyping.
Timestamps:
00:00 - Intro: Typeform context & Stack
02:00 - The Reality Gap: Hype vs. Actual Enterprise Adoption
04:31 - The Strategy: Democratizing Access & Federating Usage
06:36 - Architecture: Why we chose AWS Bedrock
08:09 - The Friction: Why IAM Roles & VPNs kill adoption
09:14 - The Unlock: Switching to Bedrock API Keys
10:04 - Internal Tooling: The "Generate Bedrock Token" CLI
11:00 - Build vs. Buy: Why AWS Bedrock beats direct Anthropic accounts
12:38 - Result: Unlimited, unmetered LLM access for every engineer
13:21 - Federated Usage: "Bring Your Own Tool" (Cursor, Emacs, CLI)
14:59 - Hiring: Replacing algo-tasks with AI-pairing interviews
17:07 - Design System: Building an MCP Server for prototyping
19:21 - Core Principles: Empowerment over Prescription
22:06 - Conclusion & Resources
✍️ Talk: Andy Kuszyk | Democratising AI in Engineering: Lessons from Typeform's Journey
🌎 This Talk was recorded at TechLead London 2025:
🔗 techleadconf.com
💫 Follow the link to watch the full version of all the conference talks, QnA’s with speakers and hands-on workshop recordings:
🔗 gitnation.com/events/techlead-conference-2025-irl
🔥 Want to hear talks like this live? Join us at TechLead Conference 2025 in London: https://techleadconf.com?utm_source=youtube&utm_medium=techlead-conference-2025-irl
TechLead Conf on X:
https://x.com/TheTechLeadConf
TechLead Conf on LinkedIn:
linkedin.com/showcase/tech-lead-conference
TechLead Conf on Instagram:
instagram.com/techleadconf
GitNation on TikTok:
tiktok.com/@gitnationorg
#TechLeadConf #aifordevelopers #GitNation #TechLeadConf2025 #WebDevelopment #TechTalks #TechLeadTalk #GitNation #modelcontextprotocol #MCP #AWSBedrock #AIEngineering #Typeform #Engineering, #LLMCostOptimization
AI adoption isn't just about giving everyone a ChatGPT Plus account. It’s about infrastructure, security, and cost 🔓
In this talk, Andy Kuszyk, Staff Engineer at Typeform, shares how Typeform built a scalable AI ecosystem using AWS Bedrock and MCP servers. Learn how to democratize LLM access while keeping your spend under control and your engineering DNA AI-driven.
Instead of prescribing specific IDEs, learn how to build a "paved road" that empowers engineers to use their preferred tools—whether it's Cursor, Cline, or Emacs—while keeping security and billing centralized.
You will learn:
🔹 Architecture: Why Typeform chose AWS Bedrock over direct vendor integrations (Anthropic/OpenAI) to solve billing chaos.
🔹 Implementation: How to replace high-friction IAM/VPN requirements with self-serve Bedrock API Keys.
🔹 Federated Usage: How to technically support any AI tool via a unified token system without vendor lock-in.
🔹 Process Impact: How unlimited AI access transforms engineering hiring interviews and design-to-code prototyping.
Timestamps:
00:00 - Intro: Typeform context & Stack
02:00 - The Reality Gap: Hype vs. Actual Enterprise Adoption
04:31 - The Strategy: Democratizing Access & Federating Usage
06:36 - Architecture: Why we chose AWS Bedrock
08:09 - The Friction: Why IAM Roles & VPNs kill adoption
09:14 - The Unlock: Switching to Bedrock API Keys
10:04 - Internal Tooling: The "Generate Bedrock Token" CLI
11:00 - Build vs. Buy: Why AWS Bedrock beats direct Anthropic accounts
12:38 - Result: Unlimited, unmetered LLM access for every engineer
13:21 - Federated Usage: "Bring Your Own Tool" (Cursor, Emacs, CLI)
14:59 - Hiring: Replacing algo-tasks with AI-pairing interviews
17:07 - Design System: Building an MCP Server for prototyping
19:21 - Core Principles: Empowerment over Prescription
22:06 - Conclusion & Resources
✍️ Talk: Andy Kuszyk | Democratising AI in Engineering: Lessons from Typeform's Journey
🌎 This Talk was recorded at TechLead London 2025:
🔗 techleadconf.com
💫 Follow the link to watch the full version of all the conference talks, QnA’s with speakers and hands-on workshop recordings:
🔗 gitnation.com/events/techlead-conference-2025-irl
🔥 Want to hear talks like this live? Join us at TechLead Conference 2025 in London: https://techleadconf.com?utm_source=youtube&utm_medium=techlead-conference-2025-irl
TechLead Conf on X:
https://x.com/TheTechLeadConf
TechLead Conf on LinkedIn:
linkedin.com/showcase/tech-lead-conference
TechLead Conf on Instagram:
instagram.com/techleadconf
GitNation on TikTok:
tiktok.com/@gitnationorg
#TechLeadConf #aifordevelopers #GitNation #TechLeadConf2025 #WebDevelopment #TechTalks #TechLeadTalk #GitNation #modelcontextprotocol #MCP #AWSBedrock #AIEngineering #Typeform #Engineering, #LLMCostOptimization
