Inspiration
We built Vatis Tech because accurate transcription for non-English languages — especially Romanian — simply didn't exist at scale. US-based tools treated European languages as an afterthought, leaving broadcasters, contact centers, and legal teams without a reliable solution. We decided to solve this from the ground up, training our own models on EU-language audio data to reach accuracy levels that finally make automated transcription production-viable.
What it does
Vatis Tech converts speech to accurate, structured text. Users can upload audio or video files (MP3, WAV, MP4, M4A, and more) or integrate via our REST API. The platform returns a full transcript with speaker labels, timestamps, and optional audio intelligence layers — sentiment, keywords, topic detection. Outputs are available as TXT, PDF, SRT, VTT, or structured JSON for API users. Use cases span call center QA, media monitoring, podcast production, medical documentation, legal transcription, and real-time broadcasting.
How I built it
Vatis Tech was built by a Romanian engineering team specializing in applied machine learning and speech science. We trained custom ASR (Automatic Speech Recognition) models on curated EU-language datasets, fine-tuned for domain-specific vocabulary and acoustic environments. The platform stack includes a scalable cloud inference layer, a Webflow-based marketing site, and a React-powered transcription editor. We have worked closely with enterprise clients to iteratively improve accuracy through feedback loops and custom model training.
Challenges I ran into
Building accurate ASR for Romanian and other EU languages meant working with far smaller training datasets than English. Achieving 99%+ accuracy required domain adaptation — training separate models for call center audio, broadcast audio, and medical dictation. Real-time transcription added latency constraints that forced infrastructure optimization. Balancing GDPR compliance with cloud scalability was an ongoing engineering and legal challenge, particularly for EU enterprise clients with strict data residency requirements.
Accomplishments that I'm proud of
Customers like Antena 3 CNN, Mediatel, AGERPRES, and Emerald Medical Center choose Vatis Tech over global alternatives — a direct validation of our accuracy and language specialization. We achieved 99%+ transcription accuracy in production environments. Our contact center vertical unlocks call QA, compliance monitoring, and agent coaching at scale. We built a partner network and serve multiple industries simultaneously — broadcasting, legal, medical, media monitoring — each with distinct accuracy and integration requirements.
What I learned
That enterprise buyers in EU markets don't just want accuracy — they need data sovereignty and local language expertise that global vendors can't credibly offer. Distribution matters as much as technology: getting into the hands of contact center operators required deep industry conversations, not just a good product. And specialization compounds — being the best Romanian speech-to-text provider opened doors to legal, medical, and broadcasting verticals in ways a generic positioning never would.
What's next for Vatis Tech
Expanded EU language coverage, deeper audio intelligence features (automated call scoring, compliance flagging, script adherence detection), and continued growth in the contact center vertical. We are also investing in real-time transcription for broadcasting and live event use cases, and building API partnerships with CRM and contact center platforms to reduce integration friction for enterprise buyers.
Log in or sign up for Devpost to join the conversation.