Senior AI/ML Engineer

Fidelity Investments•Durham, NC

6d•$107,000 - $216,000

About The Position

Note: Fidelity is not providing immigration sponsorship for this position. Principal AI/ML Platforms and Backend Engineer Bring a builder’s mindset to Fidelity’s Enterprise AI/ML Platform and help us scale the next generation of LLM-powered, agentic and search systems. You will work on the core platform that connects tools, agents, data, and models—expanding our MCP and Agentic ecosystem, abstracting vector databases, adapting and enhancing open-source libraries, and automating experimentation for high-quality retrieval and reasoning. The Team The enterprise AI/ML platform is focused on delivering AI/ML solutions for the organization. As part of this team, you will be responsible for building advanced cloud and software solutions in collaboration with Data Scientists to support packaging, deployment, and scaling of AI/ML Services in production.

Requirements

Strong Python service engineering: sound OOP, clear interfaces, thorough tests, and an obsession with readability and maintainability.
Real-world performance tuning across services and data stores: concurrency, async I/O, queues, caching, SQL/NoSQL indexing, pagination, and backpressure.
Experience building event-driven systems and/or real-time pipelines for ingestion and inference.
Mastery of debugging complex, distributed behavior—reproducible experiments, simulations, and evidence-driven conclusions.
Comfort reading open-source code and producing simplified alternatives to minimize code legacy and cognitive load.
Effective use of developer-assist tools to amplify output while keeping quality high and code bloat at minimum.
Produce services metrics that help us understand parallelism services can support in stable fashion, ensuring efficient hardware utilization. Propose scaling approaches based on application hardware utilization footprint and metrics.
Fast learning across new domains, with a knack for spotting and reducing unnecessary complexity. Team works on new products, understanding and implementing latest tech is paramount, learning fast.
Product sensibility: start from a blank slate, ask the right questions, and design primitives that feel 'Apple-like' in usability.
Produce functional picture and design of a product, based on that write requirements, epics and come up with stories which cover entire scope. Aim is that most of the risks are identified at start, not during implementation.
Collaborative communication, healthy debate, and leading by example. Comfortable switching hats to do what the projects need.
Ability to work in a highly dynamic environment
Attention to detail, being thorough in tests and questioning assumptions, being productive without a need for supervision. Our team takes pride in quality of our products.

Nice To Haves

Familiarity with key Data Science, Machine Learning, or AI libraries is a bonus, but not mandatory, as long as the candidate can demonstrate the ability to quickly learn new concepts and paradigms.
DevOps practices (CI/CD, Docker, Kubernetes) and infrastructure as code.
AWS skills: EC2, S3, RDS, Lambda, IAM etc.
Understanding Data, performant ETL, Analytics.

Responsibilities

Build the core AI/ML services running in Kubernetes and locally in 'Playground' mode
Design clean abstractions over vector databases and multistep Search/Information Retrieval pipelines
Own automated real-time data ingestion for RAG: connectors, streaming pipelines, chunking/embedding strategies, parallel processing, retrieval metrics, resilience and restartability while guaranteeing ACID integrity of processed data and elimination of redundant document processing.
Ship developer-friendly APIs/SDKs, CLIs, and templates that make it trivial to develop agents, tools, and information retrieval pipelines at enterprise scale.
Instrument everything: distributed tracing for services and agentic/tool sessions, retrieval quality metrics, performance metrics, resource usage and failure forensics.
Turn rapid prototypes into resilient systems—pragmatic designs that are simple to use, which scale in hardware efficient manner, and above all as simple as possible.
Read and distill open-source frameworks, keep what’s valuable, replace the bloated with lean, well engineered Python modules.