Project
Earnings Call Transcript Automation
Automated web scraping system extracting financial data and earnings transcripts with OpenAI integration for content generation.
→ Automated financial content pipeline
Starting point
Turning earnings calls into short posts took hours of reading and rewriting by hand.
What I built
A pipeline that pulls transcripts, summarizes them, and drafts posts on a schedule.
What changed
Publishable posts within hours of a call, at a fraction of the old effort.
The person who scoped this is the person who built it. Same person on the handoff — that's the trade you make hiring one consultant instead of an agency.
Starting point
Where things stood
Financial content teams spend hours reading earnings call transcripts, extracting key points, and summarizing them for social media distribution. The process is manual, slow, and doesn't scale across the number of companies worth covering.
Approach
The path we chose
A content pipeline that turns raw earnings call transcripts into summarized, publishable financial intelligence — automatically.
Build
How I built it
- Built a transcript monitor that scrapes MarketBeat for new earnings calls, parses the HTML into speaker-message pairs, cleans filler, and stores both raw and optimized versions.
- Sent optimized transcripts through GPT-4 with a structured prompt returning sentiment, one-sentence guidance, a short summary of metrics and risks, and key pros and cons.
- Added an optional module to post summaries to Twitter/X, supporting manual, single, or batch workflows.
The honest part
What made it tricky
The parts that didn't go to plan — and how I worked around them.
- Parsing MarketBeat required custom handling for nested Q&A sections.
- Cleaning transcripts without losing meaning was a balancing act — too aggressive hurt quality, too light kept costs high.
- Managing Twitter API rate limits and auth during batch runs needed solid error handling.
After launch
What changed
- Lowered GPT-4 processing costs by trimming transcript size.
- Fully automated from scrape to post, cutting manual effort to near zero.
- Consistent, publishable summaries within minutes of a transcript release.
Capabilities
What it does
Stack
Tools used
Services
Related services
This project is an example of these lanes — most real work touches more than one.
More projects
Adjacent work from the portfolio — same proof standard.
Project
Custodian Transition
Recovered the source data, generated thousands of documents, and routed them for e-signature end to end.
Full write-up →Project
Will It Flow
A web app with live data, cash-flow tools, and market pages that update themselves.
Full write-up →Project
WhatConverts → Salesforce
Automated handoff into Salesforce with cleanup and routing so leads arrive ready to call.
Full write-up →Need something like this?
Send a short note about what you need done — or book fifteen minutes if you'd rather talk it through.