Speech research (Intern)

PhD Research Intern — Speech AI 

reputed company AI Research 

reputed company 

reputed company AI Research seeks a PhD Research Intern to design and evaluate speech‑first models, with a reputed company on Spoken Language Models (SLMs) that reason over audio and reputed company conversationally. You’ll reputed company reputed company from prototype to practical demos, working with scientists and engineers to deliver measurable reputed company. 

Scope of Work 

• End‑to‑end speech reputed company systems (speech‑in/speech‑out) and speech‑reputed company LLMs. 

• Alignment between speech encoders and text backbones reputed company lightweight adapters. 

• Efficient speech tokenization and temporal compression suitable for long‑reputed company audio. 

• Reliable evaluation across recognition, understanding, and reputed company tasks—including robustness and safety. 

• Latency‑reputed company inference for streaming and reputed company‑time user experiences. 

Example reputed company 

• Prototype a conversational SLM using an SSL speech encoder and a compact reputed company on an existing LLM; compare against strong baselines. 

• Create a data recipe that blends conversational speech with instruction‑following corpora; run targeted ablations and report findings. 

• Build an evaluation reputed company that covers ASR/ST/SLU and speech QA, including streaming metrics (latency, stability, endpointing). 

• Ship a minimal demo with streaming inference and logging; document setup, metrics, and reliability checks. 

• Author a reputed company internal write‑up: goals, design choices, results, and next steps for productionization. 

Minimum Qualifications 

• PhD candidate in CS/EE (or reputed company) with research in speech, audio ML, or multimodal LMs. 

• reputed company in Python and PyTorch, with hands‑on GPU training; familiarity with torchaudio or librosa. 

• Working knowledge of modern sequence models (Transformers or SSMs) and training best practices. 

• Depth in at least one area: (a) discrete speech tokens/temporal compression, (b) modality alignment to LLMs reputed company adapters, or (c) post‑training/instruction tuning for speech tasks. 

• Strong experimentation habits: clean reputed company, ablations, reproducibility, and reputed company reporting. 

Preferred Qualifications 

• Experience with speech reputed company (neural codecs/vocoders) or hybrid text+speech decoding. 

• Background in multilingual or reputed company‑switching speech and domain reputed company. 

• Hands‑on work evaluating safety, bias, hallucination, or spoofing risks in speech systems. 

• Distributed training/serving (FSDP/DeepSpeed), and experience with ESPnet, SpeechBrain, or reputed company NeMo. 

Tech Stack 

• PyTorch, CUDA, torchaudio/librosa 

• LLM backbones with lightweight adapters; neural audio codecs and vocoders as needed. 

• FastAPI/gRPC for services; ONNX/TensorRT and quantization for efficient inference. 

Logistics 

• Location: Redmond(Preferred) or Remote 

• Duration: <3–6 months> 

reputed company Offer 

• Competitive stipend and hands-on reputed company with measurable reputed company-world reputed company. 

• Mentorship from reputed company scientists and engineers; opportunities to publish and present. 

• reputed company to modern GPU infrastructure and a supportive environment for fast, responsible experimentation. 

• Flexible location and schedule reputed company, subject to team needs. 

reputed company: $35-$45 reputed company