NoBroker - Manager - Data Science (8-12 yrs)

The Role:

The Manager of Data Science is a vital "player-coach" role designed for a seasoned technical expert passionate about cutting-edge LLM development, speech intelligence, and people leadership. In this role, you will lead our applied research and development efforts across both text and voice AI, establish high benchmarks for model performance, and pioneer novel techniques for voice-enabled agentic systems.

You will manage a dedicated team of data scientists and speech engineers, providing technical mentorship while remaining hands-on with the architecture of core AI capabilitiesspecifically bridging the gap between advanced LLMs and production-grade Speech-to-Text (STT) and Text-to-Speech (TTS) systems.

What You'll Do:

- Team Leadership & Mentorship: Direct and manage a team of data scientists and speech engineers, fostering a culture of scientific rigor, innovation, and continuous delivery. Mentorship is a key part of your day-to-day.

- Lead Speech & Audio R&D: Drive core research in Direct Speech-to-Speech architectures, advanced STT (automatic speech recognition, diarization, noise robustness), and highly natural TTS (voice cloning, prosody control, emotional speech synthesis).

- Advance GenAI & Agentic Workflows: Oversee projects to explore, prototype, and implement cutting-edge techniques in LLMs. Focus areas include advanced RAG strategies, model fine-tuning, multi-agent orchestration, and seamless voice-to-agent tool execution.

- Evaluation & Benchmarking: Design, build, and maintain a robust evaluation framework to rigorously benchmark our voice and text AI models against internal and external standards. Define key metrics for WER (Word Error Rate), audio latency, hallucination reduction, token cost, and guardrail safety.

- Model Stack Strategy: Help drive the strategy for our multimodal stack, determining when to use proprietary models (e.g., GPT-4o, Claude 3), open-source foundational models (e.g., Whisper, Kokoro, Llama), or custom fine-tuned specialist models.

- Cross-Functional Partnership: Act as a core technical consultant to the Product team to help define a feasible and ambitious roadmap. Partner with Engineering to develop MLOps pipelines for deploying low-latency voice bots and speech models in high-throughput production environments.

- Stay Research-Aligned: Stay constantly apprised of the latest academic research (e.g., from Interspeech, ICASSP, ACL, NeurIPS) and apply relevant findings directly to our platform.

Who You Are (Qualifications):

- Education: Bachelors, Masters, or Ph.D. in Computer Science, AI, NLP, Speech Processing, Machine Learning, or a related quantitative field.

Experience:

1. 8+ years of hands-on industry experience in data science.

2. At least 4+ years focused specifically on NLP, Speech/Audio Processing, and Machine Learning.

3. Prior experience leading small project teams or mentoring junior data scientists/engineers is required.

Speech & Audio Stack Expertise:

1. Deep understanding of STT pipelines (e.g., Whisper, Conformer, Kaldi) and voice analytics (speaker diarization, emotion recognition, VAD).

2. Hands-on experience with modern TTS architectures (e.g., Bark, VITS, XTTS, Flow-based models) and direct speech-to-speech research.

- Generative AI Stack Expertise: Demonstrable, deep, hands-on experience with the modern LLM stack, including:

1. Building and optimizing complex RAG (Retrieval-Augmented Generation) pipelines.

2. Practical experience with model fine-tuning (e.g., using Hugging Face) and parameter-efficient tuning (P)EFT techniques.

3. Deep understanding of agentic frameworks (e.g., LangChain, LlamaIndex) and tool-use mechanics.

- Technical Skillset:

1. Expert-level proficiency in Python.

2. Deep experience with modern ML/NLP/Audio libraries (e.g., PyTorch, Hugging Face transformers, diffusers, torchaudio, LangChain).

3. Experience with vector databases (e.g., Pinecone, Weaviate, Milvus).

- Execution Style: A proven ability to run structured experiments, work effectively with ambiguity, and translate complex audio/text research into practical, low-latency, high-impact enterprise solutions.

Bonus Points:

- A strong record of publications in top-tier AI, ML, NLP, or Speech conferences (e.g., Interspeech, ICASSP, NeurIPS, ACL, EMNLP).

- Experience building and deploying real-time streaming audio models or WebRTC-integrated speech pipelines.

- Familiarity with building voice agents optimized for Indian languages and accents.

Why Join ConvoZen.AI?:

- Define the Voice of AI: You won't just use the platform; you will build its core text and audio intelligence. Your work will directly define how the next generation of voice agents interact with humans.

- Work on the Frontier: You'll be solving novel, complex problems at the absolute cutting edge of technologyno two days will be the same.

- Build the Future: You will directly influence the product roadmap of a next-generation AI platform, turning your speech research into features used by world-class companies.