TaylorMohney
Senior Software Engineer
Shipping production AI and ML products, from agentic developer tooling to large-scale model serving
- TypeScript
- Python
- Go
- React
- Kubernetes
- AWS
- Years experience
- 10+Years experience
- Companies
- 7Companies
- AWS certifications
- 4AWS certifications
- Papers & preprints
- 3Papers & preprints
Professional Experience
Ten years building production software across defense, government, fintech, consumer hardware and early-stage startups.
Principal Engineer
Kindly Robotics • Remote
Building agentic developer tooling for robotics engineers, including an LLM-powered IDE and a multi-agent robot configuration framework.
Key Achievements:
- Kindly IDE, an agentic IDE for robotics engineers: shipped an LLM-powered development environment purpose-built for robotics, with chat-over-codebase awareness of URDF, ROS launch files and calibration packs, inline config completions, and an agent mode that scaffolds and validates whole robot configurations end to end
- Architected a planner/generator/validator agent graph that turns a natural-language spec (e.g. "6-DOF arm with gripper for pick-and-place at station 4") into a validated configuration covering URDF, launch files and calibration, cycled through a simulation-in-the-loop verifier before the engineer ever sees it, turning days of manual config into minutes
- Built a RAG pipeline over internal design docs, past customer configs and manufacturer specs so IDE suggestions are grounded in what the team has actually shipped rather than what the base model guessed, with citation surfacing so engineers can trust but verify
- Set up a production eval harness that replays real engineer tasks against each model version, scoring compile-pass rate, simulation success and engineer override rate, catching regressions before they reach users
- Partnered with defense contractors to embed ML pipelines for predictive maintenance, anomaly detection and adaptive process control into existing production workflows, enabling real-time operational visibility and lower defect rates at scale
Impact: Robot configuration work that took days of manual effort now lands in minutes
Technologies:
Lead Full Stack Developer
Department of Defense • Remote
Developed high-performance peer-to-peer video conferencing applications and secure data processing systems for defense applications.
Key Achievements:
- Developed a peer-to-peer video conferencing application using WebRTC, React, and TypeScript, achieving 99.9% uptime in production
- Engineered a custom signaling server in Go to manage multi-user video chat coordination, supporting seamless communication for up to 100 concurrent users
- Designed a scalable mesh network topology to support N-to-N peer connections, reducing connection latency by 30% and improving overall video call quality
- Implemented WebSocket for real-time communication and ICE protocol for NAT traversal, increasing connection success rates by 25%
- Spearheaded the development of a secure RESTful API using Node.js and Express, facilitating data exchange for a sensitive data processing application with an encryption layer that enhanced data confidentiality by 50%
Impact: 99.9% uptime in production with 30% latency reduction and 50% enhanced data confidentiality
Technologies:
Senior Full Stack Developer - Angular, React, Node, Python
USDA • Remote, Missouri
Led the design and implementation of scalable, data-driven applications using modern web technologies and cloud infrastructure.
Key Achievements:
- Led the design and implementation of scalable, data-driven applications, reducing software bugs by 35% and accelerating feature delivery timelines by 20%
- Designed and developed microservices using Angular and FastAPI, integrating with AWS Lambda and API Gateway to enable seamless communication between architectural components
- Streamlined backend processes by developing Python-based APIs, enhancing system performance and reliability, resulting in a 25% reduction in latency for key workflows
- Migrated critical features to a modular architecture using Angular, React, and AWS services, improving maintainability and deployment efficiency by 40%
- Engineered end-to-end web solutions leveraging Angular, React, Node, and Python, leading to a 50% increase in user engagement and a 30% boost in application responsiveness
Impact: 35% reduction in software bugs, 20% faster feature delivery, and 50% increase in user engagement
Technologies:
Full Stack/Web3 Developer – Solidity, React, Node, AWS
Coinbase • Remote, New York
Engineered Web3 decentralized applications and NFT marketplace solutions with advanced security features and blockchain integration.
Key Achievements:
- Engineered a Web3 dApp with advanced security features, performing in-depth mempool analysis to ensure platform scalability and reliability for high transaction volumes
- Directed a team in developing and launching a cutting-edge NFT marketplace using React, Web3.js, and Golang, reducing time-to-market by 30%
- Strategized and implemented features that drove a 20% increase in quarterly revenue, enhancing user engagement and marketplace transactions
- Architected and deployed smart contracts on Ethereum blockchain using Solidity, leveraging AWS infrastructure to handle over 500,000 transactions monthly, ensuring robust security and scalability for decentralized applications
- Configured GraphQL schemas coupled with Ethereum node API, executed through React to streamline data querying processes and optimized server communication
Impact: 30% reduction in time-to-market, 20% increase in quarterly revenue, handling 500,000+ transactions monthly
Technologies:
Contract Software Engineer
Apple • Remote
Architected and deployed end-to-end machine learning pipelines and chatbot workflows with advanced AI integration capabilities.
Key Achievements:
- Architected and deployed end-to-end machine learning pipelines using open-source tools (e.g., LangChain, Hugging Face, Ray, MLflow), supporting integration of foundation models like OpenAI and Anthropic for real-time inference and fine-tuning
- Built modular chatbot workflows with dynamic node-based logic using GraphQL and Node.js, enabling LLM agents to interact with complex data flows and external APIs in real time
- Developed CI/CD pipelines for ML models and data services using open-source DevOps stacks (e.g., Docker, GitHub Actions, Terraform, Prometheus), reducing deployment friction and model rollback time by 40%
- Designed scalable orchestration patterns that bridge private LLMs and cloud-hosted endpoints (e.g., vLLM, Ollama, SageMaker endpoints), improving security and latency across hybrid deployments
- Collaborated closely with ML engineers, frontend developers, and data ops teams to streamline model serving and system observability—cutting integration delays by over 30%
Impact: 40% reduction in deployment friction and rollback time, 30% reduction in integration delays
Technologies:
Full Stack Developer
Ai-Blockchain • Remote, California
Designed and implemented scalable blockchain-based payment systems with enhanced security and operational efficiency.
Key Achievements:
- Designed and implemented a scalable architecture for blockchain-based payment systems, enhancing operational efficiency and reducing downtime by 40%
- Conducted rigorous code reviews and established best practices for secure smart contract development, preventing vulnerabilities and enhancing code quality
- Collaborated with cross-functional teams to integrate blockchain technology with existing payment gateways, streamlining adoption for businesses
- Enhanced system reliability through automated testing pipelines, achieving a 90% reduction in critical bugs post-deployment
- Spearheaded adoption of emerging blockchain consensus protocols, improving transaction validation speed and network security
Impact: 40% reduction in downtime and 90% reduction in critical bugs post-deployment
Technologies:
Full Stack Developer Intern
Zenbase • Remote
Implemented rental payment platform features and developed secure smart contracts for blockchain-based meditation platform.
Key Achievements:
- Implemented a far-reaching rental payment platform feature using React and Node.js, enhancing user experience and increasing transaction efficiency by 25% within three months
- Implemented Solidity contracts for a proof of meditation platform, ensuring smart contract security and functionality using industry best practices
- Developed smart contracts in Solidity, enhancing gas efficiency through the use of bit manipulation and inline assembly, resulting in a 15% reduction in transaction costs
- Utilized modern tools to detect and address security vulnerabilities, enhancing platform security by 40%
Impact: 25% increase in transaction efficiency, 15% reduction in transaction costs, and 40% enhancement in platform security
Technologies:
Research & Writing
Independent write-ups on neural network quantization and parameter-efficient fine-tuning. Each one links to a public repository with the code and data behind it.
Preprints & Working Papers
Theoretical Analysis of Quantization Bounds in LoRA Fine-tuning: Error Propagation and Optimal Bit-width Selection
Authors: Taylor Mohney
Venue: Self-published preprint (2024)
Abstract
We present a comprehensive theoretical analysis of quantization error bounds in Low-Rank Adaptation (LoRA) fine-tuning. Our work establishes fundamental error bounds E[L(θ̂_q)] - L(θ*) ≤ Õ(√r/√N) + O(r·2^(-2b)σ_g²) and derives an optimal bit-width selection rule b* ≥ ½l...
Keywords
Research in Progress
Adaptive LoRA Placement for Efficient Large Language Model Fine-tuning
Authors: Taylor Mohney
Venue: Working draft (2024)
Abstract
This paper introduces a novel approach to adaptive placement of Low-Rank Adaptation (LoRA) modules in large language models. We develop algorithmic strategies for determining optimal layer positions for LoRA adapters based on gradient analysis and activation patterns. O...
Keywords
Parameter-Efficient Fine-tuning of Large Models: A Comprehensive Analysis
Authors: Taylor Mohney
Venue: Working draft (2024)
Abstract
We present a comprehensive analysis of parameter-efficient fine-tuning methods for large language models, examining the trade-offs between computational efficiency, memory usage, and downstream task performance. Our study compares LoRA, AdaLoRA, QLoRA, and other PEFT me...
Keywords
Professional Certifications
Industry-recognized certifications demonstrating expertise in cloud computing, machine learning, and software development practices.
AWS Machine Learning Certification - Specialty
Issued by: Amazon Web Services
Advanced certification in Machine Learning and Artificial Intelligence training and application with AWS technologies. Certified in data preparation and analysis/science with AWS technologies.
Skills Validated
AWS Certified DevOps Engineer – Professional
Issued by: Amazon Web Services Training and Certification
Professional-level certification in AWS DevOps technologies and applications, demonstrating expertise in implementing and managing continuous delivery systems.
Skills Validated
AWS Certified Developer – Associate
Issued by: Amazon Web Services Training and Certification
Certification in development with all AWS technologies, including application of cloud and serverless technologies.
Skills Validated
AWS Certified Cloud Practitioner
Issued by: Amazon Web Services Training and Certification
Foundational certification in AWS cloud technologies and their application, demonstrating understanding of AWS Cloud concepts and services.
Skills Validated
All certifications can be independently verified through the provided links. Active certifications are kept current through continuous professional development and renewal requirements.
Projects
A selection of research projects, web applications, and open-source libraries spanning machine learning optimization, parameter-efficient fine-tuning, and production systems.
Adaptive LoRA PlacementFeatured
Novel algorithmic approach for optimal placement of LoRA adapters in large language models
Key Achievements
- Reduced trainable parameters by up to 40% compared to uniform LoRA placement
- Maintained comparable or better downstream task performance across benchmarks
- Developed novel gradient analysis algorithms for layer selection
- +1 more achievements
BrandBeacon Platform DemoFeatured
Interactive demo showcasing brand marketing and customer engagement platform capabilities
Key Achievements
- Interactive platform demonstration with real-time data visualization
- Responsive design optimized for multiple device types and screen sizes
- Live analytics dashboard showcasing campaign performance metrics
- +1 more achievements
LLM Quantization Bounds ResearchFeatured
Theoretical analysis of quantization error bounds in LoRA fine-tuning with comprehensive experiments
Key Achievements
- Derived fundamental error bounds E[L(θ̂_q)] - L(θ*) ≤ Õ(√r/√N) + O(r·2^(-2b)σ_g²)
- Established optimal bit-width selection rule b* ≥ ½log₂(r) + ½log₂(N) + C
- Achieved strong theory-practice agreement (R>0.9) through systematic experiments
- +1 more achievements
Parameter-Efficient Fine-tuning AnalysisFeatured
Comprehensive comparison and analysis of PEFT methods for large language models
Key Achievements
- Comprehensive evaluation of PEFT methods across model sizes from 125M to 175B parameters
- Developed theoretical framework for method selection based on constraints
- Created practical guidelines for computational efficiency optimization
- +1 more achievements
Professional Portfolio Website
Modern, responsive portfolio website built with Next.js 15 and TypeScript
Key Achievements
- Lighthouse performance score: 95+
- WCAG 2.1 AA accessibility compliance
- Mobile-first responsive design
- +1 more achievements