I'm currently an NLP Engineer at Seasalt.ai. Seasalt is a startup - I started as an intern when the company was first founded, and am the most senior of two full-time NLP engineers.
You can try out two of the projects I have had the biggest hand in at Seasalt:
While working at Seasalt.ai, I've written several blogs about my projects - check out a few to get a better idea of what I've worked on!
SeaVoice Discord Bot
It started out as a proof of concept to see if we could connect Discord as a Twilio Flex channel: Discord & Twilio Flex: Bringing Flex Contact Center into Uncharted Territory...
But later I lead of team of capstone students to develop a unique Discord bot, which makes the platform more accessible by provide real-time transcriptions, text to speech, and downloadable audio files: SeaVoice Discord Bot: The First to Provide Downloadable Voice Recording with AI Transcriptions
After some time of having the bot publish, I did two case studies about how users are interacting with the text-to-speech and speech-to-text functionalities of our bot: Text-to-Speech on Discord: A TTS Discord Bot Case Study & Speech-to-Text on Discord: An STT Discord Bot Case Study.
You can add our bot to your Discord server!
NearMe Messaging
This project was my baby, and I had a hand in every aspect of it: the conversational AI, deployment infrastructure, Google Business Messages integration, and even the frontend design and marketing. NearMe Messaging allows business owners to user their Google Business profile to generate and deploy a tailor-made chatbot for their business in less than 5 minutes: Ultimate Guide to Enabling Chat on Google Maps to Connect With Customers (with examples and steps to enable it). You can try it out here!
Meeting Perception: Action Item prediction, Summarization, & Topic Extraction
Seasalt.ai's two flagship products provide real-time transcriptions and conversation insights for multi-lingual meeting & call center applications. I developed a custom dataset and used classification and summarization models to create a system that extracts action-items (or to-dos) from transcriptions. I also had a hand preparing the summarization and topic extraction models. And just as importantly, I optimized the NLP infrastructure and set up the CI/CD pipelines for these services so that they were production-ready: From Demo to Success: Meeting Perception
You can try out SeaMeet here.