Lead Speech AI Scientist – Multilingual ASR

Job descriptionLead Data Scientist-speech specialist (ASR)Location: Bangalore (India) Type: Full-Time | Immediate Joining Preferred CTC: Competitive ( 25-50 LPA )About YAL.aiYAL.Ai which stands for Your Alternative Life, is a revolutionary end-to-end communication and discovery platform that redefines how people connect, interact, and collaborate. Powered by advanced AI, YAL.ai offers secure, AI-driven instant chats, group discussions, dynamic communities, and seamless VoIP calls, delivering a comprehensive and engaging communication experience.Built on a robust Zero Trust Architecture, YAL.ai ensures maximum security by verifying every interaction and safeguarding user data at all levels. Our platform incorporates on-device AI models for enhanced privacy, zero-day fraud detection, multilingual automatic speech recognition (ASR), personalized recommendations, and comprehensive multilingual support.YAL.Ai embodies the essence of trust and innovation with its tagline, “Where AI Meets Integrity.” Users can build meaningful connections, foster vibrant communities, and explore limitless opportunities in a secure, privacy-centric digital environment. YAL.ai isn’t just a platform—it’s your alternative life, where communication and discovery are smarter, safer, and more impactful.Role OverviewWe’re seeking a highly skilled Lead Speech Researcher with deep expertise in speech technologies and advanced NLP to join our growing AI research team. In this role, you’ll work on building and optimizing machine learning systems that power intelligent audio and language-based solutions, contributing to next-generation AI products that prioritize privacy, performance, and scalability.Key ResponsibilitiesBuild a real-time voice-to-text (ASR) pipeline using models like Whisper, wav2vec2, DeepSpeech, or custom speech models. Design and implement fraud detection logic based on transcribed speech, keywords, and intent patterns. Integrate telecom call metadata to generate a multi-signal fraud score. Optimize entire model pipeline for low-latency mobile inference (TFLite, ONNX, quantization). Collaborate with VoIP/backend engineers to analyze call behavior patterns. Contribute to future capabilities: audio fingerprinting, real-time call classification, and voice anomaly detection. Work closely with AI Product and MLOps teams for deployment, updates, and feedback-based iteration.Required Technical SkillsExperience with speech AI/ASR models like Whisper, wav2vec2, DeepSpeech, Kaldi, or Silero including fine-tuning for Indian voice patterns. Strong understanding of NLP techniques for fraud detection including fraud keyword spotting, masked text decoding, and intent classification. Ability to extract and analyze telecom metadata . Hands-on experience optimizing models using TFLite, ONNX, and applying quantization (post-training or QAT) for on-device inference. Proficient in Python, with deep experience in PyTorch or TensorFlow, and comfort working with NumPy, pandas, and real-time data pipelines. (Bonus) Exposure to VoIP protocols (SIP, RTP) or experience detecting audio tampering, speaker changes, or spoofing in call audio.QualificationsBachelor’s or Master’s degree in Computer Science, Electrical Engineering, Data Science, or a related technical field. Specialization or thesis work in speech processing, ASR, telecom analytics, or applied NLP preferred. Strong portfolio showcasing real-world speech/NLP applications, open-source contributions, or peer-reviewed publications. Alumni of top institutions (e.g., IITs, IIIT-H, IISc, BITS) or equivalent global research programs are highly preferred.Experience3 to 6+ years of hands-on experience in speech AI, NLP for fraud or intent detection, or telecom-related machine learning. Previous work in R&D/product roles at communication companies or telecom/VoIP-focused startups is a major plus. Experience building, deploying, and optimizing mobile-first ML models or inference engines for real-time applications.Bonus if You HaveContributed to open-source projects like openai/whisper, mozilla/DeepSpeech, or Facebook’s wav2vec2. Kaggle competition experience in speech or fraud detection, or published work in fraud defense / voice AI. Experience working on AI security, adversarial robustness, or real-time edge inference protection.How to ApplyEasy apply on Linkedin or DM us here on LinkedIn or send your CV + work samples (GitHub, papers, demos) to hire.ai@yal.chat with subject line Subject: [Lead Speech Researcher | Your Name]