Senior AI Engineer - Data Scientist

Scry AI; Bengaluru; On-site; INR 3,500,000 - 5,200,000; Full time

Job summary

SCRY AI is seeking a Senior AI Engineer - Data Scientist with hands-on GenAI, LLM/SLM fine-tuning, Agentic AI, ASR, TTS, and Speech AI expertise. The role emphasizes turning experiments into production, with strong Python, PyTorch, Docker, and cloud/on‑prem deployment skills.

You will design and deploy production-grade GenAI and Agentic AI systems, build tool-call capabilities, and steer multilingual, low-latency speech pipelines to meet enterprise needs.

Qualifications

• Strong proficiency in Python, PyTorch, Hugging Face Transformers, and modern deep learning frameworks.
• Hands-on experience with LLM/SLM fine-tuning, model optimization, and production deployment.
• Experience building Agentic AI applications, tool/function calling, workflow orchestration, or MCP-based integrations.
• Experience developing, fine-tuning, and deploying ASR and/or TTS models.

Responsibilities

• Design, develop, fine-tune, and optimize LLMs/SLMs, ASR, and TTS models using techniques such as SFT, LoRA/QLoRA, PEFT, quantization, and knowledge distillation.
• Build and deploy production-ready GenAI, Agentic AI, Speech AI, ASR, and TTS solutions, taking prototypes from POC to scalable production systems.
• Design Agentic AI workflows involving reasoning, planning, multi-step task execution, memory, context management, and workflow orchestration.
• Build reliable tool/function-calling systems enabling LLMs to interact with APIs, databases, enterprise applications, and external services.

Skills

Python

PyTorch

Hugging Face Transformers

Deep learning frameworks

Docker

FastAPI

Git

SQL/PostgreSQL

Tools

Docker

FastAPI

Git

Job description

SCRY AI is an innovative AI-driven technology company focused on building intelligent, scalable, and high-performance solutions. We work with modern technologies across Software Engineering, AI/ML, and Data to solve real-world business problems and deliver impactful digital products.

Role Overview

We are seeking a Senior AI Engineer - Data Scientist with strong hands-on experience in Generative AI, LLM/SLM fine-tuning, Agentic AI, ASR, TTS, and Speech AI. The candidate should have experience taking AI models and solutions from experimentation and fine-tuning through production deployment, with strong expertise in Python, PyTorch, Docker, and cloud/on-premises environments.

Key Responsibilities

• Design, develop, fine-tune, and optimize LLMs/SLMs, ASR, and TTS models using techniques such as SFT, LoRA/QLoRA, PEFT, quantization, and knowledge distillation.
• Build and deploy production-ready GenAI, Agentic AI, Speech AI, ASR, and TTS solutions, taking prototypes from POC to scalable production systems.
• Design Agentic AI workflows involving reasoning, planning, multi-step task execution, memory, context management, and workflow orchestration.
• Build reliable tool/function-calling systems enabling LLMs to interact with APIs, databases, enterprise applications, and external services.
• Develop and integrate MCP-based tools and connectors, including tool schemas, parameter validation, authentication, execution, retries, fallbacks, and error handling.
• Design multi-agent and tool-use workflows with appropriate guardrails, authorization, human-in-the-loop controls, and validation for sensitive operations.
• Build and optimize multilingual and real-time speech processing pipelines, focusing on latency, throughput, memory, GPU utilization, and Real-Time Factor (RTF).
• Apply knowledge of Transformers, Conformers, CTC, RNN-T, diffusion models, neural vocoders, and speech foundation models, along with DSP concepts such as STFT/FFT, MFCCs, Mel-spectrograms, VAD, and audio preprocessing.
• Build RAG and Agentic AI applications using modern LLM frameworks and integrate them with enterprise data and systems.
• Develop backend services and APIs using Python/FastAPI and containerize solutions using Docker.
• Implement observability and evaluation for AI and agentic systems, including tool-call traces, task completion, response quality, latency, errors, reliability, and cost.
• Optimize models and AI workflows for latency, throughput, memory, scalability, GPU utilization, and production reliability.
• Collaborate with Engineering, Product, and Infrastructure teams and mentor junior team members.
• Stay current with advancements in Generative AI, Agentic AI, and Speech AI and apply relevant research and techniques to product development.

Key Qualifications

• 4+ years of experience in Data Science, Machine Learning, GenAI, or Speech AI.
• Strong proficiency in Python, PyTorch, Hugging Face Transformers, and modern deep learning frameworks.
• Hands-on experience with LLM/SLM fine-tuning, model optimization, and production deployment.
• Experience building Agentic AI applications, tool/function calling, workflow orchestration, or MCP-based integrations.
• Experience developing, fine-tuning, and deploying ASR and/or TTS models.
• Strong understanding of speech processing, DSP, audio pipelines, and speech-to-text/text-to-speech systems.
• Experience evaluating and optimizing models using metrics such as WER, CER, latency, throughput, and RTF.
• Experience with frameworks such as Hugging Face, NVIDIA NeMo, ESPnet, Kaldi, Coqui TTS, or equivalent.
• Experience with RAG, Agentic AI, LangChain/LlamaIndex, or similar frameworks.
• Strong experience with Docker, FastAPI/Python APIs, Git, and cloud or on-premise deployment.
• Working knowledge of SQL/PostgreSQL and software engineering best practices.
• Strong communication, problem-solving, ownership, and mentoring skills.

Good to Have

• Experience with Whisper, Conformers, diffusion models, VITS, HiFi-GAN, BigVGAN, or similar architectures.
• Experience building multilingual, low-latency, real-time Speech AI applications.
• Experience with GPU optimization, quantization, distributed inference/training, or model serving.
• Experience implementing AI evaluation frameworks, agent observability, guardrails, or automated testing.
• Exposure to research from Interspeech, ICASSP, NeurIPS, ICML, or ICLR.