AI Research Engineer – Speech/ASR

12+ years

voice-first healthcare AI platform

• Multilingual

  Speech-to-Text (STT)

  engine
• Healthcare-fine-tuned

  LLMs

• Large, curated

  clinical dataset

  collected over 14 years
• AI tools for clinical documentation, voice search, triage, patient interaction & more

Role: Research Engineer – Speech/ASR + LLM

1–2 years of hands-on research experience

• Speech recognition (ASR)
• Whisper/Wav2Vec/XLS-R
• Audio processing
• LLM model training / fine-tuning
• NLP model development

OR experience from:

• Research labs (IIT, IISc, BITS, etc.)
• AI startups
• ML research internships
• Academic thesis projects that resulted in real ASR/LLM outputs

Responsibilities

🗣️ Speech / ASR

• Train & fine-tune multilingual ASR models
• (Whisper, NeMo Conformer, Wav2Vec2, XLS-R)
• Build & evaluate STT models for Indian languages
• Improve accuracy for code-mixing (Hinglish/Tanglish/etc.)
• Handle real-world noisy OPD/clinic audio
• Implement chunked streaming inference
• Work with VAD, simple diarization, and segmentation
• Build WER/CER evaluation pipelines

🧠 LLM / NLP

• Fine-tune LLMs (Llama, Mistral, Gemma) on clinical datasets
• Build models for:
• Clinical summarization
• Entity extraction (symptoms, diagnosis, plan)
• Multilingual documentation
• Voice command agents
• Build training datasets from OPD transcripts
• Optimize inference (vLLM/TGI)

🏗️ Research + Engineering

• Read & implement research papers
• Prototype new architectures
• Run experiments at scale
• Collaborate with Speech + LLM teams
• Prepare reports, benchmarks & internal publications

Required Skills

Must-Have

• 1–2 years hands-on experience in ASR or LLMs
• Strong Python & PyTorch
• Experience fine-tuning Whisper / Wav2Vec / XLS-R OR LLMs
• Experience with GPU training
• Good understanding of speech features (MFCC, spectrograms)
• Experience using HuggingFace or NeMo
• Strong debugging skills & experiment tracking

Nice-to-Have

• Experience with Indian languages
• VAD, diarization (basic)
• Triton/vLLM
• Distributed training basics
• DSP fundamentals
• Research publications or internship at ML labs

How to Apply

Send your resume + GitHub + projects to:

bharat@neuralbits.com

“Research Engineer – Speech/LLM – Your Name”

Include any:

• ASR demos
• Whisper/Wav2Vec fine-tuned models
• LLM fine-tuning projects
• Research papers / thesis / experiments

Job Location : Mumbai

Salary : very competitive