研究工程师, Post-Training

PromptCube; New York (NY); Remote; USD 178,000 - 268,000; Full time

Job summary

Deepgram 正在美国远程、旧金山或纽约市招聘深度学习后训练研究工程师。该岗位将负责定位语音 AI 模型的不足、设计评估方案、准备数据并对模型进行微调，确保改进在数周内投入生产。

你需要具备扎实的 Python 与 PyTorch 技能，并有将 ML 系统交付到生产的成功记录，能将复杂客户问题转化为清晰指标并推动端到端解决方案落地。

Qualifications

• 具备训练、微调和评估深度学习模型的深入经验。
• 具备扎实的 Python 和 PyTorch 技能。
• 有将机器学习系统交付到生产的记录。
• 习惯端到端负责问题，并在过程中学习所需知识。
• 能判断哪些问题重要，以及结果何时适合交付给客户。
• 能将模糊的客户问题转化为清晰指标，并愿意使用微调、新数据或外部模型等有效方法解决。
• 积极使用并试验先进 AI 工具，能快速采用新模型和新工作方式，适应快速变化、非高度指令化的工作节奏。

Responsibilities

• 找出语音 AI 模型未达标的场景，设定可衡量的改进目标，并持续跟踪生产中的改善。
• 设计评估方案，说明模型是否变得更好以及为什么更好。
• 整理和生成训练数据，并微调模型。
• 快速搭建可用系统并投入生产，在数周内将修复交付给所有客户。
• 将模型扩展到新的语言和地区，并改进 TTS 对姓名、术语和数字的发音。
• 将说话人的声音从背景噪声和串音中分离出来。
• 让客户提前使用可操控声音、克隆声音等新语音能力，并用客户反馈指导研究方法。
• 发布面向 Voice AI 用例的新基准。

Skills

Deep learning training

Python

PyTorch

Production ML

End-to-end ownership

Decision making

Job description

岗位概况

Deepgram 招聘后训练研究工程师，加入构建大规模客户问题解决引擎的团队：定位语音 AI 模型的不足，设计评估、准备数据并微调模型，快速将改进投入生产。 岗位为全职，工作地点为美国远程或旧金山、纽约市。

工作职责

• 找出语音 AI 模型未达标的场景，设定可衡量的改进目标，并持续跟踪生产中的改善。
• 设计评估方案，说明模型是否变得更好以及为什么更好。
• 整理和生成训练数据，并微调模型。
• 快速搭建可用系统并投入生产，在数周内将修复交付给所有客户。
• 将模型扩展到新的语言和地区，并改进 TTS 对姓名、术语和数字的发音。
• 将说话人的声音从背景噪声和串音中分离出来。
• 让客户提前使用可操控声音、克隆声音等新语音能力，并用客户反馈指导研究方法。
• 发布面向 Voice AI 用例的新基准。

任职要求

• 具备训练、微调和评估深度学习模型的深入经验。
• 具备扎实的 Python 和 PyTorch 技能。
• 有将机器学习系统交付到生产的记录。
• 习惯端到端负责问题，并在过程中学习所需知识。
• 能判断哪些问题重要，以及结果何时适合交付给客户。
• 能将模糊的客户问题转化为清晰指标，并愿意使用微调、新数据或外部模型等有效方法解决。
• 积极使用并试验先进 AI 工具，能快速采用新模型和新工作方式，适应快速变化、非高度指令化的工作节奏。

申请条件与限制

• 全职岗位。
• 工作地点选项为美国远程、旧金山或纽约市；其中远程岗位标注为 USA
• Remote，不等同于全球任何地点均可申请。
• 招聘信息未进一步说明居住地、工作许可、时区、语言或出差要求，申请前应按原始招聘信息核实。
• 不要求特定学位；可通过过往项目、已上线系统或开源工作展示解决难题的能力。

加分项

• 有语音转文字（STT）、文字转语音（TTS）或语音智能体经验。
• 有分布式训练经验。
• 发表过论文、贡献过开源项目或参与过公开基准。

薪酬与福利

• 年薪 $178K - $268K USD / year。

国际远程岗位，招聘地区：USA

• Remote; San Francisco, CA; New York City, NY。请核对原招聘中的居住地、工作许可、时区、语言与出差要求；远程岗位不等于全球可申请。以下保留招聘来源英文原文。

Employer-published remote role. Listed location(s): USA

• Remote; San Francisco, CA; New York City, NY. Remote does not imply worldwide eligibility. Check the original listing for residency, work authorization, time zone, language and travel requirements.

Company Overview

Deepgram is the leading platform underpinning the emerging trillion-dollar Voice AI economy, providing real-time APIs for speech-to-text (STT), text-to-speech (TTS), and building production-grade voice agents at scale. More than 200,000 developers and 1,300+ organizations build voice offerings that are ‘Powered by Deepgram’, including Twilio, Cloudflare, Sierra, Decagon, Vapi, Daily, Cresta, Granola, and Jack in the Box. Deepgram’s voice-native foundation models are accessed through cloud APIs or as self-hosted and on-premises software, with unmatched accuracy, low latency, and cost efficiency.

Backed by a recent Series C led by leading global investors and strategic partners, Deepgram has processed over 50,000 years of audio and transcribed more than 1 trillion words. There is no organization in the world that understands voice better than Deepgram.

Company Operating Rhythm

At Deepgram, we expect an AI-first mindset—AI use and comfort aren’t optional, they’re core to how we operate, innovate, and measure performance.

Every team member who works at Deepgram is expected to actively use and experiment with advanced AI tools, and even build your own into your everyday work. We measure how effectively AI is applied to deliver results, and consistent, creative use of the latest AI capabilities is key to success here. Candidates should be comfortable adopting new models and modes quickly, integrating AI into their workflows, and continuously pushing the boundaries of what these technologies can do.

Additionally, we move at the pace of AI. Change is rapid, and you can expect your day-to-day work to evolve just as quickly. This may not be the right role if you’re not excited to experiment, adapt, think on your feet, and learn constantly, or if you’re seeking something highly prescriptive with a traditional 9-to-5.

The Opportunity

We're building the engine that solves customer challenges at scale. When customers hit a limit in our voice AI, you'll measure it, fix it, and ship the fix to every customer in weeks.

Example projects:

• Expanding our models to new languages and regions
• Improving how text-to-speech pronounces names, terms, and numbers
• Isolating a speaker's voice from background noise and crosstalk
• Giving customers early access to new voice capabilities, like steerable and cloned voices, and using their feedback to guide the research methodology
• Releasing new benchmarks for Voice AI use-cases

The Role

You will thrive in this role if you enjoy:

• Turning vague customer problems into clear metrics
• Using whatever works to fix them, including fine-tuning, new data, and external models
• Shipping work that customers use within weeks

The Challenge

We are seeking researchers who:

• Own a customer problem until it measurably improves in production
• Find the fastest path to a fix, even if it's an approach no one has tried
• Can tell when a result is ready for customers

If these expectations energize you, you might be the person we need. This role takes focus on the customer's problem, creativity in approach, and a strong drive to ship.

What You'll Do

• Find where our models fall short and set targets for improvement
• Design evaluations that show whether a model is better and why
• Curate and generate data and fine-tune models
• Build working systems quickly and ship them to production

We don't require a specific degree. Show us hard problems you've solved in past projects, shipped systems, or open source work.

You May Be a Good Fit If You Have

• Deep experience training, fine-tuning, and evaluating deep learning models
• Strong Python and PyTorch skills
• A track record of shipping ML to production
• A habit of owning problems end to end and learning what you need along the way
• Good judgment about which problems matter and when a result is ready to ship

Strong Candidates May Also Have

• Experience with speech-to-text, text-to-speech, or voice agents
• Experience with distributed training
• Published papers, open source contributions, or public benchmarks

Notice: We're aware of individuals impersonating Deepgram recruiters. All legitimate Deepgram recruiting communication comes from an @deepgram.com email address. If you've received a message claiming to be Deepgram, please forward it to [e-mail].