Deepgram
Deepgram
Deepgram este o companie de inteligență artificială vocală care dezvoltă API-uri pentru conversia vorbirii în text, a textului în vorbire și pentru înțelegerea limbajului. Platforma sa îi ajută pe dezvoltatori să creeze aplicații vocale în timp real, inclusiv soluții de inteligență artificială conversațională, instrumente pentru centre de contact și fluxuri de lucru pentru transcriere medicală. Prin procesarea limbajului vorbit, cu accent pe acuratețe, viteză și eficiență din punct de vedere al costurilor, Deepgram permite organizațiilor din domeniul sănătății, logisticii, consultanței și din alte sectoare să transforme datele vocale în informații utile și experiențe interactive.

Lead Research Scientist, Voice AI Foundations

Develop foundational models for audio codecs, speech generation, and multimodal voice AI. Advance Deepgram’s real-time speech technology through efficient, scalable research.

Descriere

  • Build neural audio codecs that compress audio at very low bit rates while preserving high fidelity across large, general-audio datasets.
  • Create steerable speech-generation models for varied voices, emotional expression, multiple speakers, environmental noise, and overlapping speech.
  • Develop embeddings that separate codec latent spaces into interpretable dimensions for speaker, content, style, environment, and channel.
  • Recombine latent representations to create synthetic audio data at unprecedented scale.
  • Train multimodal speech-to-speech systems to understand diverse people and generate empathetic, natural-sounding responses.
  • Design architectures, training methods, and inference algorithms for bare-metal hardware, enabling efficient training on billion-hour datasets and real-time inference.
  • Research latent-space models to address data, scale, and cost challenges in voice AI.
  • Run controlled experiments, ablations, evaluations, stress tests, and benchmarks to assess research hypotheses.
  • Advance speech and language AI through open-source work and research publications.

Cerințe

  • Bring strong mathematical grounding in statistical learning theory, especially self-supervised and multimodal learning.
  • Have deep expertise in foundation model architectures and scaling training across multiple modalities.
  • Connect theory with implementation by deriving new mathematical formulations and implementing them efficiently.
  • Build data pipelines to process and curate massive datasets while preserving quality and diversity.
  • Design controlled experiments that isolate architectural advances and test theoretical insights.
  • Optimize models for real-world deployment, accounting for hardware constraints and efficiency techniques.
  • Show a record of open-source contributions or research publications advancing speech or language AI.
  • Identify decisive experiments that can quickly validate or disprove ideas.
  • Recognize how to scale successful proofs of concept by a factor of 100.
  • Use AI tools to automate work and expand personal impact.
  • Adapt quickly, experiment, keep learning, and work effectively in a fast-changing AI field.

Beneficii

  • Remote work arrangement.
  • AI-first environment that encourages using and experimenting with advanced AI tools.
  • Opportunity to lead foundational voice AI research with transformative potential.
  • Opportunities to contribute to open-source projects and research publications.
  • Interview recording with the AI Notetaker is optional; opting out does not affect candidacy.

Locuri de muncă similare

FHI 360

Associate Director of Procurement, Administration and Logistics

FHI 360

Leads procurement, administration, logistics, and operations for FHI 360’s country office in the Democratic Republic of the Congo. Supervises functional leads and maintains compliance with donor, government, and supply chain requirements.

Deschide
Expivia Interaction Marketing Group Inc.

Healthcare Call Center Representative

Expivia Interaction Marketing Group Inc.

Assist members with health savings and benefit accounts through phone, chat, and email. The role includes three weeks of paid, in-person training in Erie, Pennsylvania, before moving to a hybrid schedule.

Deschide