Looking for a job

Joshua Sommers

@joshuasommers

Senior software engineer designing LLM-powered, low-latency AI systems with RAG, multi-agent orchestration, and guardrails.

United States

Message

What I'm looking for

I’m looking to build reliable, compliant, low-latency LLM/AI products—designing RAG and multi-agent systems, adding evaluation/guardrails, and owning end-to-end MLOps so models perform safely in real-world usage.

I’m a Senior Software Engineer focused on LLM and AI system design, with 8+ years of experience building production-grade intelligent applications. I specialize in RAG architectures, multi-agent orchestration, and LLM evaluation/guardrails—deploying AI copilots and decisioning systems at scale for hundreds of thousands of users.

Most recently, I architected an embedded investor copilot at Fundrise, building a production RAG pipeline with PGVector and implementing strict compliance guardrails that reduced hallucinated/non-compliant responses to low single-digit percentages. I also deliver decisioning and personalization platforms—pairing uplift/churn modeling with RL-driven policy selection—and productionize ML systems with low-latency APIs, monitoring, and MLOps practices across AWS, GCP, and Azure.

Experience

Work history, roles, and key accomplishments

Current

Senior Software Engineer

Current

Fundrise

Nov 2022 - Present (3 years 7 months)

Architected and delivered an embedded LLM investor copilot enabling natural-language queries over portfolios and scenario analyses used by thousands of active investors. Built a multi-agent RAG orchestration layer with guardrails that reduced hallucinated/non-compliant responses to low single-digit percentages and developed LLM evaluation/observability to improve production reliability.

RAG Multi agent orchestration fastAPI Pgvector OpenTelemetry LLM Evaluation

Senior Software Engineer

Scratchpad

Aug 2021 - Nov 2022 (1 year 3 months)

Led development of a Sales Intelligence and Next-Best-Action platform that used ML/NLP to score and prioritize accounts, opportunities, and tasks for thousands of sales reps. Productionized low-latency inference services and experimentation frameworks, improving follow-up completion rates and accelerating pipeline updates for reps using ML-powered queues.

Next Best Action NLP PyTorch scikit learn fastAPI gRPC A B Testing MLOps

Full Stack Engineer

Solugenix

Jul 2020 - Mar 2021 (8 months)

Built an AI-powered ticket triage and SLA-risk dashboard that predicted category, urgency, time-to-resolution, and SLA breach probability directly in agents’ workflows. Developed Python/React systems with REST APIs, deployed on AWS, and implemented feedback loops so agent overrides improved training data.

TF IDF And Embeddings scikit learn PostgreSQL React Typescript AWS CI CD MLOps

Senior Software Engineer

USAA

Oct 2019 - Jul 2020 (9 months)

Worked on an intelligent document processing platform that classified document types and extracted entities (e.g., policy IDs and coverage details) from millions of PDFs and scanned images. Implemented OCR/text-cleanup pipelines and productionized containerized gRPC/REST inference with confidence/drift monitoring to route low-confidence predictions to human review.

Document Processing SpaCy OCR Pipelines gRPC Drift Monitoring

Senior Software Engineer

Cognizant

Oct 2018 - Oct 2019 (1 year)

Built churn and upsell models using engineered RFM and behavioral features from tens of millions of transaction and interaction logs. Deployed scoring pipelines on AWS/Azure and fed scores into CRM/marketing systems, measuring lift via controlled experiments.

Churn Modeling Upselling scikit learn Pyspark Feature Engineering (RFM)Airflow

Software Developer

General Motors

Jun 2017 - Oct 2018 (1 year 4 months)

Contributed to a connected-vehicle predictive maintenance platform by analyzing billions of time-series sensor records and DTC codes to predict component failures and detect anomalous behavior. Implemented and evaluated sequence models versus tree-based approaches, integrating chosen models into batch/streaming pipelines with Kafka and Spark Streaming to generate alerts and dashboards.

Time Series Models Predictive Maintenance scikit learn Pyspark Anomaly Detection Kafka