Applied AI ML Lead [Multiple Positions Available]

JPMorganChase•Columbus, OH

6d•Onsite

About The Position

This role involves engaging in cutting-edge initiatives to enhance virtual assistant capabilities by leveraging state-of-the-art Natural Language Processing (NLP), deep learning, and generative AI techniques. The position drives the development of scalable, production-grade AI solutions through experimentation with large language models (LLMs) and small language models (SLMs), including domain-specific fine-tuning and optimization. The lead will be involved in all aspects of machine learning, supporting tech and product teams, and providing expertise. Key responsibilities include owning the machine learning solution for the transaction search application, leading the build of a question-and-answer solution for customers on chase.com, and conducting thorough analysis of business needs to inform solution design. The role requires designing and executing experiments, adapting models for the finance domain, and leading AI solution development, stakeholder management, and model release management.

Requirements

Bachelor's degree in Computer Engineering, Computer Science, Information Technology, or a related field of study plus 7 years of experience in the job offered or as Applied AI ML Lead, Sr. Specialist - Data Sciences, Tech Lead III, Sr. Tech Lead - Data Sciences, Sr. Consultant, or related occupation.
OR Master's degree in Computer Engineering, Computer Science, Information Technology, or a related field of study plus 5 years of experience in the job offered or as Applied AI ML Lead, Sr. Specialist - Data Sciences, Tech Lead III, Sr. Tech Lead - Data Sciences, Sr. Consultant, or related occupation.
Utilizing Python to implement data science solutions, build scalable machine learning (ML) pipelines, and automate workflows.
Applying Supervised and Unsupervised Learning to build predictive ML models, improve decision-making, and automate labeling.
Using feature engineering to identify and select relevant features to improve ML model performance.
Leveraging Hyperparameter Optimization to enhance ML model accuracy and generalization.
Utilizing Neural Networks including Convolution Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory's (LSTM), and Transformers to build ML solutions, automate tasks, and fine-tune domain-specific language models.
Using Open Source Embedding Models including transformers (all-mpnet-base-v2) and sentence transformers (msmarco-distilbert-base-tas-b) to capture the underlying semantic and contextual relationships in text data.
Applying Tokenization, Named Entity Recognition, Semantic Search, and Topic Modeling to structure and analyze text data, improve user experience, and automate information retrieval.
Using Prompt Engineering, System Prompt Design, Retrieval-Augmented Generation (RAG), Instruction Fine-Tuning, Parameter-Efficient Fine-Tuning, Multi-adapter Architectures, Domain Adaptation model training for Banking and Financial NLP, Synthetic Data Generation for Fine-Tuning and to Enhance LLM performance.
Utilizing Dense Retrieval, Sparse Retrieval, Hybrid Search, Embedding-based Semantic Search to improve information retrieval accuracy and efficiency.
Using Precision, Recall, F1, Mean Reciprocal Rank (MRR), Normalized Discounted Cumulative Gain (NDCG), SQUAD Metrics, Exact Match, Perplexity, Multi-class and Multi-label Evaluation, Latency Profiling, Human-in-the-loop Evaluation to assess and validate ML model effectiveness and performance.
Utilizing Distributed Training using Data Parallel, Fully Sharded Data Parallel, Mixed Precision Training, Multi-GPU Scaling for LoRA (Low Rank Adapters), Fine-Tuning of SLMs (Small Language Models) to scale SLMs model training across hardware resources.
Using KV Caching, Semantic Caching, Distributed Inference, Quantization, Low-latency API Design, Scaling LLM Serving on GPU and CPU to optimize SLMs (Small Language Models) inference speed, scalability, and resource utilization.
Employing Snowflake, Databricks, and Sagemaker to manage data and ML model training.

Responsibilities

Engage in cutting-edge initiatives to enhance virtual assistant capabilities.
Leverage Natural Language Processing (NLP), deep learning, and generative AI techniques to drive innovation and improve user interaction.
Drive the development of scalable, production-grade AI solutions through experimentation with large language models (LLMs), small language models (SLMs) and domain-specific fine-tuning.
Optimize small language models (SLMs) using advanced model fine-tuning techniques.
Lead brainstorming sessions focused on NLP advancements, LLM fine-tuning strategies, and production deployment.
Be involved in all aspects of machine learning, supporting tech, product teams, and providing expertise and guidance in machine learning applications.
Perform monthly releases and model optimization contributing to the release cycle from an ML perspective to ensure continuous improvement of the assistant's capabilities.
Own the machine learning solution for the transaction search application, which includes stakeholder management, continuous optimization, and research and innovation.
Lead the build of a question-and-answer solution using advanced NLP techniques, allowing Chase customers to ask questions on the chase.com website.
Conduct thorough analysis of business needs, exploring state-of-the-art research papers, deep learning models, and generative AI methods to inform solution design.
Design and execute experiments using both large (LLM) and small language models (SLM) to enhance performance on targeted tasks.
Utilize adapted model for the finance domain and optimize training efficiency.
Lead AI Solution Development, stakeholder management, model release management, research on NLP Solutions to drive project success and deliver production-ready solutions.