I want to clarify exactly what I am trying to build.
The goal is a production-ready ASR model for Heimish / Hasidic Yiddish, with strict, word-for-word verbatim transcription in Yiddish (Hebrew) script. The model must at minimum match the quality of an existing website that already provides a similar service, and it needs to go beyond that level.
I am prepared to invest significant effort on my side in preparing high-quality, consistent transcripts according to agreed standards. However, I am only willing to pay if the final model demonstrably meets this defined benchmark.
If the model reaches the agreed level on a fixed test set, I am prepared to compensate you generously. If it does not, the project stops and no payment is made.
The final system must be production-ready and capable of transcribing audio in seconds, from short clips to long recordings.
Because of this, I am not looking for a typical freelance engagement, but for someone confident enough in their ASR expertise to take on a success-based partnership for this project.
If this aligns with how you work, please let me know what structure you would propose.
I’m very interested in collaborating on your Heimish/Hasidic Yiddish ASR project.
I have 9+ years building production-grade AI systems, including speech recognition, NLP pipelines, LLM integration, and ASR fine-tuning for enterprise-scale deployments.
Could you please share the best way to reach you—email, Discord, or another method—so we can discuss the project in more detail?
Your goal of a production-ready ASR system that can handle both short clips and long recordings really resonates with me.
I’ve built similar systems end-to-end—fine-tuning models like wav2vec 2.0 and Whisper, integrating LLM workflows for post-processing, and deploying everything as scalable microservices using FastAPI, Docker, and Kubernetes.
Because this is a results-focused project, I’d suggest structuring our work around clear milestones and measurable outcomes so we can ensure the system meets your transcription accuracy and latency requirements.
I can put together a plan with timelines and deliverables if that works for you.
Looking forward to hearing your thoughts and discussing next steps.