About Us :

NoBroker.com is the world's largest brokerage-free real estate marketplace, dedicated to transforming the real estate experience with cutting-edge products. Founded by IIT Bombay, IIT Kanpur, and IIM Ahmedabad alumni in 2014, we've raised $366M from prominent investors like General Atlantic, Tiger Global, and Saif Partners. Our comprehensive real estate ecosystem includes NoBrokerHood for community management, NoBroker HomeServices for property management, and a range of services like Legal Services, Rental Agreements, Packers & Movers, HomeLoans, Rent Payments, and more.

ConvoZen.AI :

ConvoZen is an enterprise-grade Unified Conversational Agent Platform that turns conversations into outcomes on a single AI Agent Stack. Orchestrate multilingual workflows across omnichannel agents with persistent context, governed execution, and 100% interaction visibility to surface sentiment, compliance risk, and resolution gaps so operations improve continuously at scale.

About the Role :

We are looking for a Senior Data Scientist with 3 to 5 years of hands-on experience to drive our next-generation Generative AI, Vernacular Voice, and Agentic Workflows. In this role, you will bridge the gap between complex unstructured language systems, Speech-to-Text/Text-to-Speech integration, and production-grade Agentic RAG architectures. You will lead technical design, build scalable ML/LLM pipelines, and mentor junior engineers and technical interns.

What Youll Do :

- Agentic Workflows & Advanced RAG : Architect and deploy state-of-the-art Agentic RAG pipelines using LangGraph, LangChain, and vector databases (FAISS, Pinecone). Optimize text chunking strategies (e.g., section-aware PDF parsing) to maximize retrieval performance (MRR).

- Speech & Vernacular AI : Build and fine-tune vernacular voice systems using frameworks like AI4Bharat (Bharat4AI), STT, and TTS engines. Integrate conversational memory, confidence scoring, and guardrails to minimize LLM hallucinations.

- NLP & Large Language Models : Fine-tune open-source LLMs (e.g., Llama 3) and specialized transformer architectures (RoBERTa, SpaCy, Hugging Face) using PEFT/LoRA/QLoRA techniques for domain-specific tasks.

- Production MLOps & Orchestration : Deploy robust, low-latency microservices with FastAPI, Docker, and KubeFlow pipelines capability on large-scale datasets with above 99% uptime.

- Real-Time Latency Optimization : Implement caching mechanisms (e.g., DynamoDB, Redis), model quantization, and ONNX runtime optimizations to reduce latency down to sub-50ms for real-time inference.

- Mentorship & Technical Leadership : Mentor junior data scientists and project cohorts, drive internal hackathons/R&D initiatives, and present technical outcomes to executive stakeholders.

Who You Are (Qualifications) :

Basic Requirements :

- Education : M.Tech or B.Tech/B.E. from a top-tier institution (IITs/NITs or equivalent) with a strong quantitative foundation.

- Experience : 4+ years of end-to-end data science experience building and deploying machine learning and NLP/GenAI solutions in production.

Technical Skillset :

- Generative AI & Agents : Advanced experience with LangChain, LangGraph, Multi-Agent Systems, RAG, and Vector DBs (FAISS, Pinecone).

- Speech & NLP : Hands-on experience with STT/TTS pipelines, Indic/Vernacular language models, Named Entity Recognition (NER), SpaCy, Hugging Face, and PyTorch.

- Deployment & MLOps : Expertise in building FastAPI microservices, Dockerization, KubeFlow pipelines, and deploying on AWS (SageMaker, S3, EC2, DynamoDB).

- Core ML & Python : Expert-level Python, SQL, and solid grounding in classical ML (XGBoost, Scikit-Learn) and Explainable AI (XAI).

Preferred / Bonus Qualifications :

- Experience developing open-source projects, patents, or research publications.

- Track record in high-stakes domains (FinTech, Digital Payments, Logistics, Fraud Detection).

- Background in leading technical interns, managing cross-functional deliverables, or public speaking at industry events.

The Role :

We are looking for a hands-on Senior Data Scientist with 3 to 5 years of experience to join the core AI team at ConvoZen.AI. In this role, you will work closely with our Data Science Manager and engineering teams to research, prototype, fine-tune, and deploy production-grade models across both Speech Processing (STT/TTS) and Generative AI (Agentic RAG & LLMs). If you thrive on taking complex audio/text models from research papers into real-time, low-latency production systems, this role is for you.

What Youll Do :

- Speech & Audio Engineering : Build and fine-tune Speech-to-Text (STT) and Text-to-Speech (TTS) models. Optimize speaker diarization, Voice Activity Detection (VAD), emotion detection, and noise robustness for real-world call environments.

- GenAI & Agentic Workflows : Implement advanced Retrieval-Augmented Generation (RAG) pipelines, multi-agent frameworks, and tool-use mechanics to allow real-time voice-to-agent interactions.

- Model Fine-Tuning & Quantization : Apply parameter-efficient fine-tuning ((P)EFT, LoRA/QLoRA) and quantization on open-source LLMs and speech models to optimize latency and compute costs.

- Evaluation & Benchmarking : Measure and optimize key performance indicators, including Word Error Rate (WER), Time-To-First-Audio-Packet (TTFA), hallucination rates, and guardrail compliance.

- Production MLOps Integration : Partner with platform engineering to package models via FastAPI and Docker, integrating them into high-throughput streaming audio pipelines (e.g., WebRTC/WebSockets).

- Mentorship : Help guide junior data scientists and project interns in experimental rigor, code reviews, and structured model evaluation.

Who You Are (Qualifications) :

Experience & Education :

- Education : Bachelors or Masters degree in Computer Science, AI, Speech Processing, Signal Processing, or a related quantitative field.

- Experience : 3 to 5 years of hands-on data science experience in industry, with at least 2+ years focused explicitly on NLP, Speech/Audio Processing, or Generative AI.

Technical Skillset :

- Speech & Audio Stack :

1. Hands-on experience with modern STT tools (e.g., Whisper, Faster-Whisper, Conformer, Kaldi).

2. Familiarity with TTS architectures (e.g., Bark, VITS, XTTS) or audio frameworks (torchaudio, Librosa).

3. Understanding of voice analytics (speaker diarization, VAD, audio signal processing).

- GenAI & LLM Stack :

1. Demonstrable experience building complex RAG systems and using agentic frameworks (LangChain, LangGraph, LlamaIndex).

2. Hands-on fine-tuning experience with Hugging Face transformers.

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

- Core Engineering :

1. Expert proficiency in Python and ML frameworks (PyTorch, Scikit-Learn).

2. Experience with API development (FastAPI) and containerization (Docker).

Bonus Points :

- Hands-on experience building voice agents optimized for Indian languages and regional accents (e.g., AI4Bharat / Indic STT & TTS frameworks).

- Exposure to real-time streaming audio protocols (WebRTC, WebSockets, telephony streams).

- Track record of hackathon wins, open-source contributions, or publications in top AI/speech venues (Interspeech, ICASSP, ACL, NeurIPS).