Trey Doig is a software engineer and repeat founder behind SeatMe, acquired by Yelp, and Echo AI, acquired by Calabrio. As Echo AI’s co-founder and former chief technology officer, he built systems that transform customer conversations into operational intelligence while making generative AI reliable enough for enterprise use.
From restaurant reservations to enterprise AI
Doig studied at North Carolina State University and worked as an engineer at IBM and Creative Commons before co-founding SeatMe with Alexander Kvamme. After Yelp acquired the reservations company in 2013, Doig led engineering work integrating its reservation search into Yelp’s consumer platform.
In 2017, Doig and Kvamme founded Pathlight, initially focused on performance management for customer-facing teams. As chief technology officer, Doig began developing large-language-model applications in late 2022 and helped turn operational reporting into enterprise conversation intelligence: analyzing customer interactions themselves to identify problems conventional metrics overlook. Pathlight became Echo AI, which Calabrio acquired in December 2024.
Analyze conversations beyond the support queue. Customer calls, chats, and tickets can reveal manufacturing defects, broken deliveries, cancellation drivers, and emerging dissatisfaction. Doig builds systems that examine interactions comprehensively instead of relying on small manual samples.
Make accuracy an operational discipline. Reliable analysis starts with collecting, transcribing, and normalizing customer data, then applying configurable evaluation criteria. His enterprise AI demonstration shows solution engineers reviewing summaries, correcting grades, and tracking hallucinations and model drift. Echo AI used Log10’s evaluation tooling; Log10 developed that product independently.
Design infrastructure around changing models and business needs. Doig’s account of production language-model infrastructure describes combining commercial and open models, retrieval, embeddings, self-hosting, and domain-specific tuning to manage throughput, cost, and customer-specific requirements.
Keep human judgment inside automated workflows. His writing about contact-center AI addresses privacy, hallucinations, bias, data quality, and model-selection tradeoffs. Human-in-the-loop evaluation lets domain specialists correct outputs and define quality standards, while multi-step AI agents coordinate more complex analysis across connected tasks.
Echo AI’s customer-conversation pipeline shows how broad analysis becomes useful only when teams can inspect outputs, grade them against customer needs, and turn corrections into better models.
▶ Watch ↗