Principal AI/ML Engineer (Large Language Model)

CACI•Aurora, CO

22h•Onsite

About The Position

The Principal AI/ML Engineer will support the development of AI/ML algorithms in a multitude of disciplines from object detection/classification, natural language processing, reinforcement learning, and large language models. This role involves leading and mentoring a multidisciplinary team to implement machine learning algorithms for diverse customer challenges. The engineer will apply Large Language Models (LLMs) to various applications, fine-tune foundation models, build adaptors, and implement retrieval augmented generation (RAG) techniques. They will also build custom applications with LLM frameworks, deploy LLM solutions, and analyze large multi-domain datasets to identify features for actionable data. Additionally, the role requires reviewing publications to apply cutting-edge concepts and interfacing with leadership to communicate technical status.

Requirements

BS in machine learning, computer science, mathematics, or related fields.
10+ years of experience, preferably in software development or as a data scientist with 2+ years of building LLM applications using some of the following: Fine-tuning foundational models, Steering Techniques (e.g Sparse auto encoders, representation tuning), Building adapters to use foundational models (e.g. PEFT, llama factory), Prompt engineering techniques / Inference time techniques (e.g. chain of thought, tree of thoughts, etc.), Using Retrieval Augmented Generation techniques to populate and query vector databases (e.g. Weaviate, pinecone), Using LLM Frameworks (e.g. LangChain, DSPy), Using AI APIs ( e.g AWS Bedrock, OpenAI), Using LLM deployment frameworks (eg llama.cpp, vllm, tgi), Developing UIs with ReAct
Experience leading an interdisciplinary team of researchers and software developers and working with a program manager to define project scope and schedule to ensure we meet project milestones as defined by our customers
Experience with Python and data science / machine learning libraries (e.g. PyTorch, TensorFlow, Keras, OpenCV, NumPy, Pandas, Polars, scikit-learn, etc.)
Active TS/SCI U.S. Government Security Clearance

Nice To Haves

MS or PhD in machine learning, computer science, mathematics, or related fields.
Experience leading an interdisciplinary team of researchers and software developers
Experience with any of the following Computer Vision domains: Large Language Models and experience identifying ways to incorporate them into new areas and applications
Applying Transformer-based architectures to domains in other areas outside of Natural Language Processing (NLP) such as computer vision
Object detection algorithms such as YOLO and Faster-RCNN
Natural Language Processing algorithms such as BERT
Generative Adversarial Networks and Variational Autoencoders
Reinforcement learning and familiarity with Gymnasium Gym, RLlib, and Stable Baselines
Applying clustering algorithms and/or deep neural networks to real life problems
Implementing tracking and pattern-of-life algorithms
Experience with Machine Learning libraries and frameworks such as HuggingFace and LangChain
Experience with Computer Vision libraries such as OpenCV, Nerfstudio, FiftyOne, etc.
Experience with Linux
Familiarity with using AWS cloud computing resources such as EC2, S3, Lambda, etc.
Experience with any of the following additional languages: Java, C++, Rust, Go, and/or C#
Experience implementing algorithms on the GPU in Python or C++ using CUDA and other CUDA libraries
Experience with implementing tracking and pattern-of-life algorithms
Experience in application deployment, virtualization, and containerization (e.g. Podman, Docker, Kubernetes, Rancher)
Experience working with various Remote Sensing datasets (e.g. EO/OPIR/SAR images, passive RF, etc.)
Experience shaping and writing proposals

Responsibilities

Lead and mentor a multidisciplined team consisting of developers and researchers to implement machine learning algorithms to solve a broad set of challenges for our various customers
Apply Large Language Models (LLMs) to a variety of applications within remote sensing such as tasking collections, identifying gaps in collection plans, analyzing patterns of life, and more.
Fine tune foundation models and building adaptors for new applications (llama factory, PEFT)
Apply retrieval augmented generation (RAG) techniques to data to populate and query vector databases (e.g. Weaviate)
Build custom applications with LLM frameworks such as LangChain, DSPy
Deploy LLM solutions across cloud-based and local resources using kubernetes (llama.ccp, vllm etc)
Analyze large multi-domain datasets such as images, text and/or graph data, to identify statistically relevant features to build models that provide analysts with actionable data
Review relevant publications to understand and apply cutting edge concepts to defense and commercial applications
Interface with both internal and external leadership to communicate technical status