Location Name: Pune Corporate Office - Mantri
Job Purpose
The Senior AI Engineer role exists to design, build, and operationalize production-grade AI models and pipelines, enabling scalable Voice AI and Generative AI solutions aligned with business use cases. The role focuses on hands-on development, optimization, and deployment of AI systems, translating architectural vision into robust, high-performing solutions.
Duties And Responsibilities
- Build and deploy Speech AI and LLM-based systems (STT, TTS, S2S, dialogue orchestration)
- Implement production-grade pipelines for inference, fine-tuning, and model lifecycle
- Work on low-latency, high-throughput model serving (real-time voice systems)
- Optimize models using quantization, distillation, pruning techniques
- Integrate LLMs/SLMs into voice workflows (prompting, chaining, orchestration)
- Develop emotion-aware dialogue handling logic and fallback strategies
- Support voice biometrics and anti-spoofing system implementation
- Work closely with Product and Lead AI to translate business problems into AI solutions
- Ensure model performance, monitoring, observability, and continuous improvement
- Build and convert POCs into stable production deployments (no demo-only work)
- Follow best practices in MLOps, versioning, and reproducibility
Key Decisions / Dimensions
- Model implementation choices (fine-tune vs prompt vs orchestration)
- Selection of frameworks, libraries, and deployment patterns
- Trade-offs between performance vs cost vs scalability
- Decisions on model optimization techniques (quantization, distillation, etc.)
- Integration approach for LLMs with speech pipelines
- Handling edge cases in dialogue flow and failure scenarios
Major Challenges
- Making models production-ready (latency, stability, cost) — not just proof of concept
- Handling noisy real-world voice inputs across languages and dialects
- Balancing accuracy vs latency vs infra cost constraints
- Integrating multiple AI components (STT + LLM + TTS) without breaking flow
- Managing model degradation and continuous learning loops from failures
- Working within real-world infra limitations (GPU availability, edge constraints)
Required Qualifications And Experience
- Bachelor's or Master's degree in Computer Science, AI, or related field
- Experience: 3–6 years in AI/ML with strong hands-on delivery
- Strong experience in Speech AI (STT, TTS, S2S)
- Hands-on experience with LLMs/SLMs (OpenAI, HuggingFace, LangChain)
- Experience in real-time AI systems / low-latency inference pipelines
- Proficiency in Python, PyTorch / TensorFlow
- Experience with model optimization (quantization, distillation)
- Knowledge of MLOps, deployment pipelines, and model monitoring
- Understanding of dialogue systems and conversational AI flows
- Exposure to voice biometrics / anti-spoofing (good to have)
Nice to Have
- Experience with Indic languages / dialect-heavy environments
- Hands-on work in production AI (not just research/POC)
- Exposure to edge AI / on-device deployment
