AWS Automation Developer for S3 Transcriptions

AWS Automation Developer S3 Call Recording Transcription, AI Summary, and Email Workflow We are a property management company looking to automate our call recording workflow. Our phone system, GoTo Connect, currently uploads call recordings to an AWS S3 bucket. Right now, we manually log in to AWS, download recordings, upload them elsewhere for transcription, paste that transcript to summarize them with AI, and then paste the Claude or GPT summary into our tenant notes. We want to replace that manual process with a small, reliable AWS-based workflow. Project Goal Build an automated system that detects new call recordings in S3, transcribes them, summarizes them using AI, stores the output, and emails the right people. --Desired Workflow When a new call recording is uploaded to our S3 bucket: 1. S3 trigger detects the new audio file. 2. AWS Lambda (or equivalent) starts a transcription job using Amazon Transcribe. 3. Transcript is saved to an archive folder in S3. 4. Transcript is summarized using either OpenAI API, Anthropic (preferred) or AWS Bedrock (if we have to - our team already uses GPT and Claude). 5. A call summary and transcript are emailed automatically. 6. A weekly summary email is sent to our “bosses” email address. --Summaries: each entry is identified by the phone number primarily + secondarily by name or other identifiers present in the call (the only reliable tag in S3 each time is the callers phone number) 7. Files and outputs are organized in S3 for future reference. --Each line item int he summary links back to the recording itself. ## Required Email Outputs ### 1. Per-call email Each completed call should generate an automatic email to a designated system email address. The email should include: * Call date and time, if available * Caller/callee phone number, if available from file name or metadata * Full transcript in the email body or as an attached TXT, PDF, or DOC/DOCX file. TXT preferred for compatibility on mobile, mac, pc etc. for all team members regardless of device. The end result will always be a copy/paste into our PM software notes * AI-generated call summary * Action items or follow-up items, if any * Any urgency or risk flags, such as: * maintenance emergency * rent/payment issue * legal threat * angry tenant * habitability/safety issue * complaint * lease violation * Link to the archived transcript/output in S3 + audio file We do not need the audio file attached to the email reports. ### 2. Weekly manager summary email Once per week, send a summary email to a separate “bosses” email address. The weekly email should include: * Total number of calls processed * Calls grouped by type, if possible: * maintenance * rent/payment * leasing * complaint * vendor * other * Calls with unresolved action items * Calls flagged as urgent or high-risk * Short summaries of the most important calls * Links to transcripts or archived files where appropriate ## AWS Components We are open to recommendations... * S3 trigger * AWS Lambda * Amazon Transcribe * OpenAI API / Anthropic (preferred) or AWS Bedrock (least preferred unless it integrates easy and doesn't add cost for us - we currently only pay for S3 storage) for summarization * Amazon SES (? recommend please if alternative is better) for email delivery * S3 archive folder for transcripts, summaries, and logs * EventBridge or scheduled Lambda for weekly summary email ## Deliverables Please provide: * Working AWS automation * Clean S3 folder structure for archived transcripts and summaries * Email templates for per-call and weekly summary emails * Basic error handling and logging * Setup documentation * A simple handoff guide so our team understands how it works * Any API keys/secrets stored securely, not hardcoded * Recommended retention policy for transcripts and call summaries -- Nice-to-Have Features * Detect phone numbers from filename or metadata * Tag calls by category (Maintenance, Leasing, Concern/Complaints, Other) * Create a tenant-note-ready summary format (1-3 paragraphs + bullet points easy to read format - we can provide examples of our current notes and summaries) * Create different summary formats for maintenance, leasing, rent/payment, and complaints * Ability to resend or regenerate a transcript/summary if needed * Basic dashboard or spreadsheet log of processed calls --Important Notes * Reliability matters more than tool complexity. * We do not want a fragile workflow that breaks every time a file name changes for example.. * We prefer a clean AWS-native setup vs,. disconnected no-code tools. * We need the system to be understandable and maintainable after the project is complete. * We use Claude and GPT for other functions, so would prefer to keep with one of those tools for the summary step, depending on cost, reliability, and security pros/cons. ## Ideal Candidate Experience with: * AWS S3 * AWS Lambda * Amazon Transcribe * Amazon SES * EventBridge or scheduled AWS workflows * Python or Node.js * OpenAI API and/or AWS Bedrock * Secure handling of API keys and environment variables * Building practical business automations Experience with call recording workflows, property management, transcription systems, or AI summarization is a plus. --Budget Please provide an expected range of hours to complete this. And whether you are willing to make small adjustments later if needed (1-2 hours paid follow up later if needed). We are looking for a clean, practical version one first. Draft Idea for Milestones: * Milestone 1:** Confirm architecture + access requirements * Milestone 2:** S3 → Transcribe → archive working *Milestone 3:** AI summary + per-call email working * Milestone 4:** Weekly summary email + documentation