AI Model Architect

Role Overview: As the architect and owner of Neo's AI infrastructure, you will be responsible for training custom models for unique use cases, building production ML pipelines, and creating reasoning systems to make Neo intelligent. You will work across the full ML lifecycle, from data pipelines to model training to production deployment. Key Responsibilities: - Develop specialized models that foundation models cannot provide, such as training speaker diarization for Indian accents, fine-tuning embedding models for conversational memory, and developing custom NER for Hindi-English code-mixing. - Optimize models for edge deployment, handle class imbalance in emotion detection and intent classification. - Build memory architecture and ML pipelines, including designing temporal knowledge graphs, extracting entities and relationships, and enabling longitudinal pattern detection. - Own the end-to-end speech pipeline, including training/fine-tuning ASR models for Indian languages, building speaker diarization systems, developing audio quality assessment models, and optimizing for edge deployment. - Create the query understanding and reasoning system, including building hybrid retrieval, training ranking models, developing proactive insight detection, and fine-tuning LLMs for conversational queries. - Design agent orchestration for specialized AI agents, including training classifier models, building reward models, developing action prediction models, and creating meta-learning systems. - Build enterprise ML APIs with custom model serving, design multi-tenant architecture, implement A/B testing infrastructure, and create auto-scaling inference pipelines. Qualifications Required: - 2-5 years of experience building and deploying ML/DL models in production serving real users at scale. - Robust expertise in PyTorch or TensorFlow for training, optimization, debugging, and deployment. - End-to-end ML ownership from data pipeline to model training to production monitoring and iteration. - Deep learning fundamentals including architectures, optimization, regularization. - Proficiency in Python, system design, and pragmatic approach to building ML systems. - Strong plus in speech processing or NLP, knowledge graphs, model compression, LLM fine-tuning, reinforcement learning, Indian language experience, open-source contributions, or Hugging Face ecosystem. Additional Company Details: - Greenfield ML problems with a focus on Indian accent diarization, Hindi-English entity linking, and temporal conversation understanding. - Full-stack ownership including training models, building data pipelines, optimizing for edge, deploying at scale, monitoring quality, and iterating based on metrics. - Founding team equity in a fast-growing startup defining a new category. - Work with technical founders with AI background from IIT Madras, small team, high autonomy, and real impact in building conversation intelligence. - Market timing in the nascent ambient computing space where the models built will set standards for conversational AI infrastructure. (Note: The JD does not contain any additional details about the company) Role Overview: As the architect and owner of Neo's AI infrastructure, you will be responsible for training custom models for unique use cases, building production ML pipelines, and creating reasoning systems to make Neo intelligent. You will work across the full ML lifecycle, from data pipelines to model training to production deployment. Key Responsibilities: - Develop specialized models that foundation models cannot provide, such as training speaker diarization for Indian accents, fine-tuning embedding models for conversational memory, and developing custom NER for Hindi-English code-mixing. - Optimize models for edge deployment, handle class imbalance in emotion detection and intent classification. - Build memory architecture and ML pipelines, including designing temporal knowledge graphs, extracting entities and relationships, and enabling longitudinal pattern detection. - Own the end-to-end speech pipeline, including training/fine-tuning ASR models for Indian languages, building speaker diarization systems, developing audio quality assessment models, and optimizing for edge deployment. - Create the query understanding and reasoning system, including building hybrid retrieval, training ranking models, developing proactive insight detection, and fine-tuning LLMs for conversational queries. - Design agent orchestration for specialized AI agents, including training classifier models, building reward models, developing action prediction models, and creating meta-learning systems. - Build enterprise ML APIs with custom model serving, design multi-tenant architecture, implement A/B testing infrastructure, and create auto-scaling inference pipelines. Qualifications Required: - 2-5 years of experience building and deploying ML/DL models in production serving real users at scale. - Robust expertise in PyTorch or TensorF