Software Developer/Engineer (Mid Level experience)

TechArmy•Philadelphia, PA

10d•Hybrid

About The Position

We are seeking a skilled Software Developer/Engineer with mid-level experience to focus on on-premise Large Language Model (LLM) and Vector Database implementation. This contract role requires hands-on experience deploying open-source LLMs, proficiency in Python for LLM-related tasks, and expertise in vector databases and Retrieval-Augmented Generation (RAG) pipelines. A strong understanding of data privacy, security, and enterprise requirements is essential. The role involves creating a reference architecture, a working prototype, and providing documentation and knowledge transfer to internal teams.

Requirements

Hands-on experience deploying open-source LLMs such as Meta Llama 3 and Mistral / Mixtral in on-prem or private environments.
Strong proficiency in Python for LLM inference, prompt engineering, and integration.
Experience with CPU-based inference, model quantization, and performance tuning.
Practical experience with open-source vector databases such as Qdrant, Chroma, Milvus, or pgvector.
Proven implementation of Retrieval-Augmented Generation (RAG) pipelines.
Experience generating and managing embeddings and metadata filtering.
Understanding of data privacy, air-gapped deployments, and enterprise security requirements.
Experience implementing access controls and audit logging.

Nice To Haves

Experience with LangChain or LlamaIndex.
Exposure to Rust, Go, or C++ for high-performance services.
Familiarity with Docker and Kubernetes for on-prem deployments.
Knowledge of inference frameworks (e.g., vLLM, llama.cpp, Hugging Face Transformers).
Prior work in regulated or enterprise environments.

Responsibilities

Deploying open-source LLMs such as Meta Llama 3 and Mistral / Mixtral in on-prem or private environments.
Utilizing Python for LLM inference, prompt engineering, and integration.
Optimizing LLM performance through CPU-based inference, model quantization, and performance tuning.
Working with open-source vector databases such as Qdrant, Chroma, Milvus, or pgvector.
Implementing Retrieval-Augmented Generation (RAG) pipelines.
Generating and managing embeddings and metadata filtering.
Understanding and implementing data privacy, air-gapped deployments, and enterprise security requirements.
Implementing access controls and audit logging.
Creating a reference architecture and deployment guidance.
Developing a working prototype (LLM + vector DB + RAG).
Providing documentation and knowledge transfer to internal teams.