Gawaine O'Gilvie
Senior Software Engineer
Building AI-powered products and scalable backend systems.
- Built internal platforms used by engineering teams at scale
- Developed AI-powered applications (LLMs, RAG pipelines, vision-language models)
Eighteen months of independent consulting since Sonos — client delivery across backend and frontend, alongside self-directed platform work. Currently focused on the intersection of AI systems and backend infrastructure — building production-grade tools that put language models to work at scale. Most energized by green-field projects, platform problems, and work that requires both systems thinking and AI literacy. Open to full-time roles and contract engagements where technical depth matters.
AI & ML Projects
Signal to Alpha — Options Trading Platform
Retail traders lack systematic tools to capture unstructured Discord trade callouts, execute them across multiple brokers, and analyze performance with AI-powered critique
Built a full-stack options trading platform: a 4-stage rule-based parsing pipeline (classify → extract → normalize → score) that routes low-confidence callouts to human review, a multi-broker execution layer (Schwab OAuth 2.0 + Webull HMAC-SHA1) with Fernet-encrypted token storage and idempotent order IDs, 4 interchangeable automated strategies behind a common interface with shared server-side risk guards, and AI trade critique via Claude with tiered market data, schema-validated output and TTL caching
End-to-end trade lifecycle from Discord callout to P&L analytics, tax estimation, and journal — covered by 4,800+ automated tests spanning parsing, strategy replay, broker execution, restart recovery, and database migrations; live WebSocket price streaming to all connected clients, real-time bot config reloads without restarts
DocSage — RAG Document Intelligence Platform
Teams waste hours manually reviewing large document sets with no semantic search, structured traceability, or grounded AI answers tied to real source passages
Built a production RAG pipeline: multi-format ingestion (PDF with dual-parser fallback, HTML, Markdown), 512-char sliding-window chunking, OpenAI text-embedding-3-small embeddings stored in pgvector with an HNSW index, cosine similarity retrieval via a custom SQL stored function, and GPT-4o synthesis with Pydantic-validated structured output and retry logic via tenacity — deployed on AWS RDS PostgreSQL + Render + Vercel
Live at docsage.phoenix7.dev — every response returns exact citations, token counts, and USD cost; automated coverage across ingestion, retrieval, and generation
Debugging production incidents required hours of manual log parsing to identify root causes
Built a full-stack app (Next.js + FastAPI) that analyzes production logs using OpenAI GPT-4 to identify errors, cluster patterns, and surface root causes with actionable remediation steps
85% reduction in debugging time for production incidents and CI/CD failures
Core Engineering
Developer Platform Tooling
Engineers lacked standardized workflows for testing and deployment across distributed services
Built internal tooling to streamline CI/CD, contract validation, and deployment workflows for cloud services
Reduced deployment time by 40% and eliminated 85% of contract violations across 10+ microservices
Enterprise Inventory Management Platform
A media-heavy inventory product was outgrowing proxy-based uploads, fragile authorization, and manual deployment practices
Introduced direct-to-S3 uploads with pending-state confirmation, hardened media lifecycles and admin authorization, aligned frontend API contracts, and scripted Docker deployments to AWS
Uploads up to 100 MB bypass the application servers; admin access was restored, media cleanup became deterministic, and deployments became repeatable
Experience
Freelance Consulting
Senior Full-Stack Engineer & Technical Consultant
- Sole engineer on a 14-month client engagement delivering a full-stack inventory platform across both tiers — Node/TypeScript/Express/Prisma backend and Next.js 16 / React 19 frontend — 298 commits, ~46k LOC, 71 test files, 27 database migrations
- Designed a multi-vertical variant architecture shipping one codebase as multiple domain-specific products across four variant databases; caught a silent data-loss class where unknown-key stripping made cross-variant writes return 200 while discarding the client's data, and ordered a rejection middleware ahead of validation to close it
- Re-architected an options trading engine from process-per-user to a single event-driven process — strategy runners publish immutable signals to a typed bus, per-user executor coroutines consume them — so strategy logic executes once regardless of subscriber count; shipped as 9 dependency-ordered PRs with the legacy path live throughout, halving worker memory allocation
- Integrated two brokerages across five protocols (OAuth 2.0 REST, WebSocket streaming, HMAC-SHA1 request signing, MQTT, gRPC) behind a broker-agnostic order layer, with Fernet-encrypted credentials, idempotent order IDs, and database-level delivery uniqueness so a retry cannot double-execute
- Built a real-time availability and dispatch-priority API for a municipal emergency-services client on Server-Sent Events — chosen over WebSockets for a strictly server-to-client flow — and refactored the monolith into router/controller/service layers with unit and functional suites across three versioned releases
- Deployed a production RAG platform (FastAPI + pgvector on AWS RDS + Next.js) where every generated answer cites its source chunks with similarity scores — retrieval traceability designed in as a product requirement, not retrofitted for debugging
Sonos, Inc.
Senior Software Engineer (Cloud Team)
- Migrated 10+ applications to AWS EKS, reducing infrastructure costs by 40% through optimized containerization and resource allocation
- Built Lambda/S3 data pipelines processing millions of events with robust error handling and batch processing for reliable data ingestion
- Developed automation systems that reduced Kubernetes migration time by 80% and streamlined daily deployments for cloud services
- Built developer productivity platforms using React and TypeScript, increasing engineering team satisfaction by 30%
- Mentored junior engineers on TypeScript best practices, AWS architecture patterns, and Kubernetes deployment strategies
Nuance Communications
Software Engineer
- Engineered real-time WebSocket services in Python (Flask, Bottle) handling high-throughput voice recognition traffic in production
- Built and maintained microservices and RESTful APIs supporting speech and analytics workflows across internal and external systems
- Developed monitoring dashboards that improved incident response time by 40% with real-time visibility into system health
- Collaborated with data scientists to optimize speech recognition models and improve accuracy metrics
Technical Skills
Languages
Backend & Systems
AI / ML
Cloud & Infrastructure
Frontend
Databases & Observability
Education
Master of Science, Data Science
University of Texas at Austin
Austin, TX
- Advanced coursework in Machine Learning, Statistical Analysis, and Data Engineering
- Specialization in predictive modeling and big data analytics
Bachelor of Science, Computer Engineering
Northeastern University
Boston, MA
- Focus on Software Engineering and Systems Architecture
- Coursework: Data Structures, Algorithms, Embedded Systems, Computer Networks
Work With Me
Have a complex AI or backend problem?
I help teams turn technically demanding ideas into reliable production systems. Available for contract and consulting engagements.
- Build AI-powered applications (LLMs, RAG pipelines, intelligent automation)
- Design scalable backend systems and cloud infrastructure
- Improve developer velocity, CI/CD, and platform tooling
Prefer LinkedIn? Connect with me there.
Available for full-time roles and contract / consulting work · Remote or hybrid