Miko - Lead NLP/Artificial Intelligence Engineer - Python/Deep Learning

The Role

We are looking for an experienced Lead AI NLP Engineer in the team. In this role, you will be leading the efforts in the research, design, development, and optimization of state-of-the-art natural language processing algorithms and models for creating a robust conversational experience on the Miko series of robots. We expect you to have strong coding, mathematical, and implementation skills with expertise in Python deep learning frameworks.

Work Experience & Qualifications

5 years or higher industry experience in the NLP domain. No restrictions on degree/qualifications; only skills matter.

Responsibilities

• Spearhead the developmental efforts in the NLP domain such as improvisation of existing conversational stack, design new features, and develop state-of-the-art algorithms for robust language understanding and conversational experience.
• Supervise and develop production ready frameworks for applied AI algorithms in diverse applications such as language context understanding models, open domain question answering, large scale semantic search systems as well as robot personality modules.
• Design, implement and optimize large scale data processing and continuous training pipelines for various NLP tasks on a single/distributed GPU system to accelerate model training development cycles.
• Consistent coordination and collaboration with the AI backend, Content, Linguistics and QA team for efficient development cycles and testing for new features and improvements.
• Ensuring timely releases, maintenance and scaling of services in the production environment.

• Excellent understanding and familiarity with the Pytorch framework.
• Proficiency with Transformers architecture based models.
• Working experience with Large scale question answering/information retrieval systems.
• Experience with at least one of the large scale dense vector search libraries such as FAISS, SCANN, ANNOY, NGT.
• Experience working with Large language models (LLMs) and prompt engineering
• Experience with large scale Intent prediction/Entity extraction systems.
• Proficiency with distributed training and inference techniques in Pytorch.
• Thorough understanding of model compression/optimization/quantization techniques for faster GPU inference.
• Excellent understanding of lower level fundamentals of deep learning models such as layer details, attention, normalization, backpropagation, and optimizers.
• Experience in transfer learning and fine-tuning of BERT models as well as designing architectures from scratch for downstream tasks.
• Strong understanding of CPU-GPU transfers, CUDA usage, GPU architectures, concepts of pipelining and multiprocessing as well as latency/throughput bottlenecks during training of neural networks.
• Ability to read and implement related academic papers in the NLP domain.
• Proficiency in machine learning libraries/frameworks such as Gensim, Sklearn, Spacy, Flair, Stanza, Allen-NLP, Sentence Transformers.
• Proficiency in one of the databases: MySQL, OrientDB, MongoDB, PostgreSQL
• Experience in deployment of NLP models at scale on a GPU enabled system using one of the following frameworks: Torchserve, TFServing, Triton, KFServing, BentoML.
• Experience with one of the libraries for model inference via API: Flask/Falcon/FastAPI.
• Experience in data augmentation and synthetic generation techniques in NLP.
• Proficiency with Linux OS and basic bash scripting.

(ref:hirist.tech)