Joseph is available for hire
Hire JosephJoseph OBrien
Verified Expert in Engineering
Software Engineer and AI Developer
Philadelphia, PA, United States
Toptal member since August 6, 2025
Bio
Joseph O'Brien is a platform and developer tooling engineer with deep experience across Rust- and Python-based systems, building reliable internal tools for software organizations. He also maintains a substantial open-source tooling portfolio across agent workflows and developer productivity systems. He brings a pragmatic, systems-level approach to complex environments: clear failure modes, strong test coverage, better operational visibility, and maintainable code that teams can extend safely.
Portfolio
Toptal
Agentic AI, Rust, Large Language Models (LLMs), AI Agents...
Level 3 Labs
Python, AIOps, Multistage LLM Chains, Machine Learning, Azure, LangChain...
Freelance Clients
Python, TypeScript, JavaScript, Large Language Models (LLMs), SQL, AI Agents...
Experience
- Python - 8 years
- FastAPI - 5 years
- API Development - 5 years
- Retrieval-augmented Generation (RAG) - 4 years
- Generative Artificial Intelligence (GenAI) - 4 years
- Agentic AI - 3 years
- LangChain - 3 years
- Multistage LLM Chains - 3 years
Preferred Environment
Linux, Helix, MacOS, Zed, Warp
The most amazing...
...work I’ve done is helping teams turn fragile infrastructure and AI workflows into reliable systems they could scale with confidence.
Work Experience
Senior AI/ML Developer
2025 - 2026
Toptal
- Partnered on production engineering efforts spanning developer platforms, cloud-integrated services, AI/ML infrastructure, and workflow automation.
- Built reliable data and model-adjacent systems for structured pipelines, retrieval and search, embeddings, and application-facing AI workflows.
- Strengthened production systems through better diagnostics, validation, testing, release hardening, and operational safeguards.
- Contributed across the full delivery path from architecture and implementation through CI/CD, observability, and production integration.
- Brought a systems-engineering approach to ML/AI-adjacent work by emphasizing validation, reproducibility, testing, and maintainable architecture over one-off experimentation.
Technologies: Agentic AI, Rust, Large Language Models (LLMs), AI Agents, Large Language Model Operations (LLMOps), Claude, Anthropic, Google Cloud Platform (GCP), Amazon Web Services (AWS), CI/CD Pipelines, Testing, Machine Learning Operations (MLOps), PyCharm, Bash, Bash Script, Claude Code, Data Quality, JSON, Multi-agent Systems, Project Management, Software Architecture, System Design, Regression, Agentic RAG Systems, Claude API, Workflow Automation, RAG Architecture, Asynchronous Programming, Redis, Architecture, Maintainability, Multi-agent Orchestration, Performance, Reliability, Scalability, APIs, Amazon Virtual Private Cloud (VPC), Firewalls, Graphics Processing Unit (GPU), Identity & Access Management (IAM), Terraform, Docker
Co-founder, Developer
2024 - 2025
Level 3 Labs
- Spearheaded the development of cutting-edge AI applications by leveraging frameworks such as LangChain, LangGraph, and PydanticAI to deliver innovative, agentic solutions for SME clients.
- Collaborated directly with clients to translate business needs into tailored technical specifications and end-to-end, production-grade software solutions.
- Led a distributed team of developers, ensuring timely project delivery, high code quality, and exceptional client satisfaction.
- Engineered robust RESTful APIs, enabling seamless integration with 3rd-party platforms and enhancing product capabilities.
Technologies: Python, AIOps, Multistage LLM Chains, Machine Learning, Azure, LangChain, API Development, FastAPI, CI/CD Pipelines, Ollama, OpenAI, LangGraph, Agentic AI, AI Agents, Gemini, Prompt Engineering, Retrieval-augmented Generation (RAG), Large Language Models (LLMs), AI Chatbots, Full-stack Development, Data Engineering, Microsoft Azure, Vector Databases, AI Prompts, Graph Databases, PG Vector, Gradient Boosting, LightGBM, XGBoost, Artificial Intelligence (AI), AI Data Classification, Cloud Run, Google Cloud Platform (GCP), API Design, Optical Character Recognition (OCR), PDF, Data Science, Natural Language Processing (NLP), AI-generated Code, GitOps, User Experience (UX), User Interface (UI), Cloud, Technical Leadership, Recommendation Systems, Amazon Web Services (AWS), Machine Learning Operations (MLOps), Back-end, Object-oriented Programming (OOP), Chatbot Conversation Design, Chatbots, Conversational AI, SOP Development, Conversational Agent, Predictive Modeling, Web Scraping, NVIDIA TensorRT, Image Processing, NVIDIA CUDA, Object Detection, Deep Learning, Multimodal GenAI, Multimodal Models, Legal Technology (Legaltech), Full-stack, PostgreSQL, OpenAI API, REST APIs, Azure Functions, Fine-tuning, Zapier, n8n, Vector Search, API Integration, Front-end Development, AI Design, AI Programming, LoRa, System Architecture, Agentic Frameworks, CrewAI, System Architecture Design, OpenAI GPT-4 API, Automation, Data Scraping, Scraping, Website Data Scraping, Communication, Team Mentoring, Supabase, Next.js, Management, Data Pipelines, Natural Language Queries, Claude, Cursor AI, PyCharm, Bash, Bash Script, Visual Studio Code (VS Code), Data Quality, JSON, PySpark, Leadership, Multi-agent Systems, Project Management, Software Architecture, Startups, Team Leadership, System Design, Beautiful Soup, Regression, Natural Language Toolkit (NLTK), SpaCy, Agentic RAG Systems, Workflow Automation, RAG Architecture, Asynchronous Programming, Email Automation, Redis, Architecture, Maintainability, Multi-agent Orchestration, Performance, PyTorch, Reliability, Scalability, APIs, Firewalls, Graphics Processing Unit (GPU), Identity & Access Management (IAM), Docker
Software Developer
2018 - 2024
Freelance Clients
- Delivered custom software and LLM-powered solutions to diverse SMB clients, improving operations, automation, and business workflows.
- Consulted on cloud-based architecture, ensuring scalability, reliability, and modern best practices in AI and full-stack development.
- Worked cross-functionally with designers, data engineers, and client IT staff to deliver projects exceeding expectations and industry standards.
Technologies: Python, TypeScript, JavaScript, Large Language Models (LLMs), SQL, AI Agents, Prompt Engineering, Retrieval-augmented Generation (RAG), Computer Vision, AI Chatbots, Full-stack Development, Vector Databases, AI Prompts, Artificial Intelligence (AI), AI Data Classification, API Design, Optical Character Recognition (OCR), PDF, Data Science, Natural Language Processing (NLP), AI-generated Code, GitOps, User Experience (UX), User Interface (UI), Cloud, Technical Leadership, Recommendation Systems, Back-end, Object-oriented Programming (OOP), Chatbot Conversation Design, Chatbots, Conversational AI, SOP Development, Conversational Agent, Predictive Modeling, Web Scraping, Image Processing, Deep Learning, Multimodal GenAI, Multimodal Models, Full-stack, AutoGen, PostgreSQL, OpenAI API, REST APIs, Azure Functions, Zapier, n8n, Vector Search, API Integration, Front-end Development, AI Design, AI Programming, LoRa, System Architecture, Agentic Frameworks, System Architecture Design, OpenAI GPT-4 API, Automation, Data Scraping, Scraping, Website Data Scraping, Communication, Team Mentoring, Supabase, Next.js, Management, Data Pipelines, Natural Language Queries, Bash, Bash Script, Visual Studio Code (VS Code), Data Quality, JSON, Leadership, Project Management, Software Architecture, Team Leadership, System Design, Beautiful Soup, Natural Language Toolkit (NLTK), SpaCy, Workflow Automation, Asynchronous Programming, Architecture, Performance, PyTorch, Reliability, Scalability, APIs, Firewalls, Identity & Access Management (IAM), Docker
Owner, Developer
2022 - 2023
StreamShift
- Acquired and managed a digital business providing custom web and streaming overlays for Twitch, YouTube, and Facebook, scaling client engagement through targeted social marketing.
- Designed and maintained the company's full-stack web application (MERN stack: MongoDB, Express.js, React, Node.js), delivering high-availability products to streamers.
- Hired and directed a marketing team to launch multi-channel campaigns, resulting in increased brand exposure and revenue.
Technologies: JavaScript, EJS, SCSS, HTML, MongoDB, Heroku, React, Node.js, MERN Stack, Express.js, Full-stack Development, Data Engineering, User Experience (UX), User Interface (UI), Cloud, Technical Leadership, Back-end, Object-oriented Programming (OOP), Full-stack, REST APIs, API Integration, Front-end Development, System Architecture, System Architecture Design, Automation, Communication, Team Mentoring, Management, Data Pipelines, JSON, Leadership, Project Management, Software Architecture, Startups, Team Leadership, Beautiful Soup, Asynchronous Programming, APIs
Experience
Internal Tooling
A Rust-based application with a proprietary framework. Researched and tested leading frameworks. Designed and built AI Agents, MCP servers, and data classification components for training large language models (LLMs) and maintained CI/CD and testing workflows.
Minibox: Agent-controllable Container Runtime
https://github.com/89jobrien/miniboxI built an open-source, agent-controllable container runtime in Rust with a daemon and CLI architecture. Minibox pulls OCI images and provides Linux namespace isolation, cgroups v2 limits, and overlay filesystem support. Its hexagonal architecture keeps runtime adapters swappable at startup. Linux and GKE are production-supported, while macOS uses experimental VM-backed adapters with limited execution and logging. The workspace includes conformance, property, protocol end-to-end, system, and benchmark tooling.
Crux: Agentic Rust DSL and Task Runtime
https://github.com/89jobrien/cruxBuilt and evolved an open-source Rust system for agentic workflows, task execution, and structured runtime behavior. My work included adding pipeline step handlers, expanding the CLI surface, implementing a SQLite-backed registry back end, and improving overall reliability with integration tests and property-based validation around dependency graph behavior. I also refactored larger modules into clearer boundaries and removed dead code to keep the system maintainable as the architecture grew. The project reflects my interest in building durable automation primitives rather than one-off scripts: typed workflows, explicit contracts, testable behavior, and tooling that supports more complex, higher-level agent systems.
Obfsck: Open-source Secrets Detection and Compliance Scanner
https://github.com/89jobrien/obfsckDeveloped and maintained an open-source secrets detection and compliance tool designed for real engineering workflows rather than idealized demos. My work included expanding regression coverage for secret-pattern edge cases, improving allowlisting and repo-local configuration, and adding stronger CI safeguards against unsafe runtime behavior. I also used targeted refactoring and test design to improve reliability and code health over time. The project sits at the boundary of security, developer tooling, and workflow automation: catching sensitive data risks early while remaining practical enough to integrate into everyday commit and review pipelines.
Taskit: Config-driven CI and Workflow Orchestration
https://github.com/89jobrien/taskitCreated and expanded an open-source workflow and CI orchestration tool focused on safer release processes and better engineering ergonomics. I added automated promotion and finish flows, conflict detection and resolution hooks, richer error modeling, and integration tests covering real failure modes such as dirty worktrees, wrong-branch execution, and dry-run behavior. I also improved operator-facing output so users could understand step-by-step results and failures more clearly. The core goal of the project was to turn brittle manual release steps into repeatable, testable automation that engineers could trust across active repositories.
Agentlint: Linter for AI Coding Agent Harnesses
https://github.com/89jobrien/agentlintDeveloped an open-source linter for AI coding-agent harness files spanning multiple ecosystems, including Claude Code, Cursor, Codex, Gemini, OpenCode, and Pi. My contributions focused on expanding validator coverage, improving heuristics around frontmatter and command/config patterns, and strengthening confidence through property-based and integration testing. I also improved reporting and smart project traversal so the tool could work effectively in real developer environments rather than only in narrow examples. The project is about bringing the same rigor to AI-agent configuration that engineers already expect from linters and static analysis in traditional codebases.
StllAI
Full-stack application developed from the ground up to provide valuation and analysis for spirit barrels. It serves collectors, investors, and enthusiasts by offering deep financial insights and investment recommendations.
1Tool
https://allbots.ioDeveloped exclusively for AllBots, 1tool serves as the robust, foundational toolkit that underpins the platform’s comprehensive suite of AI agents. These agents were meticulously designed and developed to encompass a wide range of tasks, such as voice recognition, automated email responses, and customer service workflows.
RechtshilfeAI
Developed a compliant, offline-capable AI platform for German legal professionals, offering instant access to validated resources and automated document workflows. I led model integration, privacy architecture, and UX development.
FareboxRAG
This project entailed the development of a RAG-based artificial intelligence (AI) assistant tailored for Canadian transit maintenance operations. The primary objectives of this project were to provide real-time information accessibility, automate workflows, and offer operational support to mobile teams. Heim Enterprises was the designated contractor for this project.
Looprs: LLM Agent Runtime
https://github.com/89jobrien/looprsI built a Rust runtime and CLI for multi-turn LLM agent sessions with tool use and multiple model providers. The system separates domain types, provider adapters, tool execution, and terminal interfaces so agent behavior remains testable as capabilities expand.
Taskit
https://github.com/89jobrien/taskitI built config-driven CI and workflow orchestration in Rust. Added affected-crate detection, protocol-drift tracking, promotion and finish flows, CI gates, conflict handling, dry runs, and guards for dirty worktrees and invalid branches. Designed explicit error models and operator-facing step results, backed by integration tests for real failure paths. Taskit replaces fragile release checklists with repeatable automation.
Agentlint
https://github.com/89jobrien/agentlintI built a Rust linter for AI coding-agent harnesses across Claude Code, Cursor, Codex, OpenCode, Gemini, Pi, and Looprs. It validates agents, skills, commands, hooks, settings, MCP configuration, instruction files, and documentation frontmatter. The Cargo workspace uses a thin CLI with focused validator crates, supports human-readable and JSON output, and integrates with CI. Its test suite includes unit, regression, and partial property testing.
mcpipe: Universal API-to-CLI Adapter
https://github.com/89jobrien/mcpipeI built mcpipe to turn MCP servers, OpenAPI specifications, GraphQL endpoints, and existing shell tools into a consistent command-line interface. It supports MCP over stdio and HTTP/SSE, OpenAPI 3.x JSON or YAML, GraphQL introspection, and help-text parsing for existing CLIs. It also provides source discovery, schema caching, authentication and custom headers, JSON formatting and filtering, result limiting, command search, and OpenAPI generation from CLI backends.
Langchainx: Rust Framework for LLM and Agent Systems
https://github.com/89jobrien/langchainxI maintained and expanded a production-ready fork of langchain-rust to build composable LLM applications in Rust. The fork is organized into 15 focused crates with typed errors across chain, tool, loader, and provider boundaries. It supports agents, retrieval workflows, vector stores, semantic routing, structured output, and source-code loaders for 11 languages. I also established three levels of end-to-end testing using deterministic fakes, local Ollama, and real Postgres and Qdrant services, all running in lightweight VMs without Docker.
Updog: Trace-driven Agent Improvement Loop
https://github.com/89jobrien/updogI built a systematic, trace-driven agent improvement loop. Its CLI collects shell-command traces, combines human and LLM feedback, generates Promptfoo evaluation templates, ranks feedback clusters with HALO scoring, and produces agent-improvement handoff documents. The seven-phase workflow also supports a gated, optional automation heartbeat and validates the harness before a commit and archive checklist. Its Rust workspace separates domain types from CLI adapters, and custom trace collectors implement a dedicated port.
Certifications
DECEMBER 2024 - PRESENT
Certified AI Developer
IBM
Skills
Libraries/APIs
OpenAI API, Pydantic, REST APIs, Beautiful Soup, Claude API, API Development, React, Node.js, NumPy, XGBoost, PySpark, Natural Language Toolkit (NLTK), SpaCy, PyTorch, EJS, OpenAPI
Tools
PyCharm, Claude, Claude Code, Helix, AI Prompts, Zapier, n8n, Amazon Virtual Private Cloud (VPC), Terraform, Azure OpenAI Service, Logging, LMMS, Zed
Languages
Python, SQL, JavaScript, TypeScript, Bash, Bash Script, SCSS, HTML, GraphQL, Rust
Frameworks
LangGraph, Agentic Frameworks, Flask, Chainlit, Streamlit, LightGBM, AutoGen, Next.js, Express.js
Paradigms
Object-oriented Programming (OOP), Automation, Management, Asynchronous Programming, Maintainability, Testing, DevOps, Model Context Protocol (MCP)
Platforms
Linux, Ollama, Azure, Docker, Google Cloud Platform (GCP), Amazon Web Services (AWS), Azure Functions, CrewAI, Replit, Heroku, MacOS, Visual Studio Code (VS Code), Cloud Run, NVIDIA CUDA
Storage
Data Pipelines, JSON, PostgreSQL, Graph Databases, Redis, MongoDB
Industry Expertise
Project Management
Other
AIOps, Multistage LLM Chains, Machine Learning, LangChain, Large Language Models (LLMs), Web App Development, Natural Language Processing (NLP), Sentiment Analysis, Data Science, Generative Artificial Intelligence (GenAI), Agentic AI, AI Agents, Retrieval-augmented Generation (RAG), Pgvector, Artificial Intelligence (AI), Full-stack Development, Optical Character Recognition (OCR), Technical Leadership, Back-end, Conversational Agent, Web Scraping, Deep Learning, Full-stack, API Integration, AI Design, AI Programming, OpenAI GPT-4 API, Data Scraping, Scraping, Website Data Scraping, Communication, Team Mentoring, Cursor AI, Data Quality, Leadership, Multi-agent Systems, Software Architecture, Startups, Team Leadership, System Design, Regression, Agentic RAG Systems, Workflow Automation, RAG Architecture, Multi-agent Orchestration, Performance, FastAPI, CI/CD Pipelines, OpenAI, Hugging Face, Prompt Engineering, Qdrant, Pinecone, Gemini, AI Chatbots, Data Engineering, Microsoft Azure, Vector Databases, PG Vector, Gradient Boosting, AI Data Classification, API Design, PDF, AI-generated Code, GitOps, User Experience (UX), User Interface (UI), Cloud, Recommendation Systems, Machine Learning Operations (MLOps), Chatbot Conversation Design, Chatbots, Conversational AI, SOP Development, Predictive Modeling, Image Processing, Multimodal GenAI, Multimodal Models, Fine-tuning, Vector Search, Front-end Development, LoRa, System Architecture, System Architecture Design, Supabase, Natural Language Queries, Email Automation, Architecture, Reliability, Scalability, APIs, Firewalls, Graphics Processing Unit (GPU), Identity & Access Management (IAM), Computer Vision, MERN Stack, Structured Logging, Command-line Interface (CLI), AI Assistants, Open-source Software (OSS), NVIDIA TensorRT, Object Detection, Legal Technology (Legaltech), SaaS, Platform as a Service (PaaS), Large Language Model Operations (LLMOps), Anthropic, AI Security, Static Analysis, AI Compliance, Containerization, Virtualization, Infrastructure, Systems Engineering, Cloud Infrastructure, Productivity, Task Flows, Property-based Testing, Release Management, Workflow Automation & System Integration, Systems Design, Model Validation, Integration Testing, Validation, Warp, Linting
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring