Model Research Engineer (m/f/d)

Location: Remote / Berlin (Office available) | Language Requirement: Fluent German At Deepslate, we are building Speech to Speech Voice AI models that sound and act indistinguishable from a human. And we believe everyone should be able to use it. When it comes to text and images, giants like OpenAI and Google have already cracked the code. With video, Veo3, Sora and others are closing the gap rapidly. But with its endless languages, dialects, accents, subtle intonations, and speech melodies voice remains a highly complex unsolved frontier. That is exactly why we started Deepslate. Backed by top-tier investors from the Tech and AI sectors, as well as a major German VC fund, we are incredibly well-funded and moving fast. We are building the future of communication. We aren't trying to build another standalone platform; instead, we are the intelligence engine powering countless other applications. Whether it's integrated as a module by a CRM provider, plugged into another Voice AI platform, or directly embedded into an enterprise system by our integration partners - our model is everywhere. Your Role Our proprietary Voice AI model is the core of Deepslate. Your mission is to research, train, and refine this very model. We aren't just stitching together existing APIs; we are building a speech to speech model that understands and replicates genuine human nuances. You will tackle complex, unsolved challenges like real-time emotion recognition, subtle intonations, and natural speech melodies to build a top-tier voice model from the ground up. What You'll Do: Your work sits directly at the intersection of AI research and scalable software engineering. You will define how natural, fast, and reliable our models perform in production. Research & Train: Design, train, and evaluate our proprietary deep learning models. You will drive the development of new capabilities, such as emotion recognition, and integrate them into our core architecture. High-Performance ML Systems: Optimize our models for maximum inference speed and efficiency, ensuring they can handle massive datasets and real-time workloads at scale. Software Engineering: Write clean, production-ready code. We need researchers who are also strong software engineers, capable of translating complex research into highly available architectures. MLOps & Tooling: Build and maintain the internal pipelines for data processing, model training, and evaluation to make our iteration cycles fast and reliable. What We’re Looking For: Must-Haves: Deep expertise in Python and modern deep learning frameworks (e.g., NeMo, PyTorch , or similar). Proven hands-on experience working on high-performance, large-scale ML systems, particularly in the context of Language Modeling (text or audio). A true startup mindset: You are comfortable navigating the glorious chaos of an early-stage codebase (inconsistent structures, missing docs. You name it, you’ll help fix it). Drive and Vision: You aren't just looking to blindly process Jira tickets. You want to actively shape the product, architect solutions, and take complete project ownership. We are VC-backed and growing fast. You should be highly motivated by the prospect of joining a startup at an early stage and scaling alongside us. Nice-to-Haves: Previous experience in a fast-paced Startup or DeepTech environment. Hands-on experience with audio processing, Speech-to-Text (STT), Text-to-Speech (TTS), or multimodal models. Interested in experimenting with inference engines in rust in the future Fluent German skills (spoken and written). What We Offer: True R&D Freedom: We give you the time, resources, and trust to research and build technology that simply doesn't exist yet. You won't be optimizing wrappers around other APIs; you will be solving fundamental, uncracked challenges in Voice AI. Competitive Compensation: We are well-funded by top-tier VCs. We offer a highly competitive salary. We are looking for top talent, and we are ready to pay for it. Serious Compute: You will have access to the hardware and cloud infrastructure necessary to actually train and test large-scale models without constantly hitting bottlenecks. Absolute Flexibility: Fully remote work, asynchronous communication, and zero micro-management. As outlined in our "How We Work" section, we care strictly about your output and impact, not your schedule. Deepslate is an equal opportunity employer. We welcome applications from all qualified candidates regardless of gender, nationality, ethnic origin, religion, disability, age, or sexual orientation.