Performance & reliability
Rails, PostgreSQL, Sidekiq, Redis and Kubernetes — traced across the stack, not tuned in isolation.
Slow endpoints · queue backlogs · OOMKills · database pressureSenior Rails & AI Systems Consultancy
Find the bottleneck. Fix the system. Keep shipping.
We diagnose production Rails systems, implement the fix, and stay through rollout — from PostgreSQL, Sidekiq, Redis and Kubernetes to AI workflows and product development.
Discuss a challengeConsulting + Engineering
Start with the problem, not a staffing plan.
Bring us a production bottleneck, a risky technical decision, or a product that needs senior ownership. We investigate, recommend and implement.
AI systems
Grounded LLM workflows, agent integrations and automation designed for production constraints.
Architecture · evaluation · guardrails · implementationProduct & platform engineering
Senior delivery for new products, integrations and business-critical features from prototype through rollout.
Rails · React · APIs · data infrastructureRescue & modernization
Independent diagnosis and hands-on recovery for legacy systems, risky upgrades and stalled projects.
Codebase audits · migrations · observability · architectureWays to work with us
A small first step. A clear next decision.
Consulting does not end with a slide deck. We can diagnose the system, implement the fix and stay through production verification.
- 01
Diagnostic sprint
Usually one focused week of read-only discovery, telemetry review and technical analysis. You leave with evidence, priorities and a decision-ready readout.
- 02
Remediation sprint
We implement the highest-value fixes, add regression protection, measure the change and plan a safe rollout and rollback.
- 03
Embedded engineering
Ongoing senior support for teams that need architecture, incident follow-through and hands-on delivery without a long lock-in.
- 04
Product development
Discovery, prototyping and production development for products and features that need both product judgment and engineering depth.
Selected Work
Selected Work
Client engagements, product experience and R&D across Rails systems, infrastructure and practical AI.
Development
Napster: Web Backend Development and R&D for a Music Streaming Pioneer
We contributed to Napster's web backend platform and conducted research and development for a new product initiative, supporting one of the most recognizable names in digital music.
See project
Development
TwoMargins: Crowdsourcing Financial Document Analysis for Wall Street
We built TwoMargins — a financial document annotation platform featured in the Wall Street Journal — handling complex SEC document parsing with Nokogiri and supporting real-time collaborative analysis. Bloomberg used the platform to publish their annotated edition of Warren Buffett's 2015 shareholder letter.
See project
Development
Roomsie: From Failed Builds to Revenue in 8 Months
Roomsie is a housing marketplace for interns and students in Washington, D.C. After multiple development teams failed to deliver, we took over, rewrote the application from scratch, and launched a revenue-generating product in 8 months.
See project
Production-safe by default
Advice that survives contact with the system.
Evidence before prescriptions
We baseline the system and separate symptoms from root causes before recommending work.
Read-only first
Discovery starts with the least access needed. Production changes require an explicit plan and approval.
Scoped decisions
Each phase has a defined question, deliverable and stop gate. No open-ended transformation programme.
Implementation included
When the recommendation is clear, we can write the code, tune the infrastructure and own the rollout.
Receipts, not promises
Tests, benchmarks, screenshots, diffs and rollback plans back every claim about what changed.
Direct senior access
You work with the engineers doing the analysis and implementation. Your repositories and assets remain yours.
Open Source
Open source R&D
We fine-tune LLMs on production Rails code and release the weights. Run them locally with Ollama.
31BGGUF
qwen3-coder-30b-rails
Production-grade Rails code generation. MoE architecture — only 3B params active per token.
View on Hugging Face8BGGUF
qwen3-8b-rails
Lightweight model that runs on a laptop. 5 GB download, 8 GB RAM minimum.
View on Hugging Face7BLoRA
qwen2.5-coder-7b-rails
LoRA adapter for Qwen 2.5 Coder 7B. Merge with base model for Rails-tuned generation.
View on Hugging Face3BLoRA
qwen2.5-coder-3b-rails
Compact LoRA adapter. Fast inference for code completion and small generation tasks.
View on Hugging Face