$ ls ./services/
Services
I've been doing applied AI for years, long before ChatGPT. That means decisions backed by numbers, not hype. If your problem is solved by a regression, I'll tell you. If you need an agent, I'll ship it to production ready to scale and with cost under control.
$ ./ai-agents
AI agents consulting
We decide together whether your problem needs an agent, a fine-tuned model, a classical classifier or nothing new. Then I take it to production with hard metrics tied to P&L.
What's included
- Architecture choices driven by data: when an agent, when retrieval, when a smaller model, when just rules.
- Working POC in a few weeks. Code quality and documentation ready for your team (current or next) to take it to product without losing context.
- Per-response cost treated as a first-class metric: caching, batching, model routing, audited prompts.
- Quantitative evaluation pipelines on your own datasets so you iterate on outcomes, not intuition.
- Guardrails, tooling, memory and observability built for real traffic.
$ ./mentoring
1:1 and team technical mentoring
Mentoring for engineers and tech leads who want to ship real AI, not demos. Founding member of AI teams and 50,000+ students trained.
What's included
- Recurring sessions with measurable goals and concrete deliverables per sprint.
- Code and AI architecture reviews, not theoretical classes.
- Career coaching for senior engineers and tech leads moving into AI leadership.
- In-house programs for teams that need to level up fast in applied AI and architecture.
$ ./digital-transformation
AI-driven digital transformation
AI adoption with judgment. I map use cases by real ROI, not by what's trending on LinkedIn this week.
What's included
- Opportunity audit: which cases justify generative AI, which need classical analytics, which need nothing new.
- Roadmap prioritized by P&L impact, with measurable milestones, realistic budget and team shape.
- Hands-on work with your team to ship the first product in weeks instead of camping in POC forever.
- Governance, security and metrics plan so the adoption survives the next stack or model swap.
$ ./platform-architecture
Web platform architecture
I review and redesign your platform so it scales, is cheap to operate and supports fast iteration. With experience taking products past 5 million users.
What's included
- Latency, infrastructure cost and tech-debt audit with an action plan prioritized by impact.
- Defensible stack decisions: Python, Rust, Cloudflare Workers, AWS, vector databases only when they pay back.
- Microservices, queues, search and data layer designed to take growth without yearly rewrites.
- Refactor of hot services in production without stopping product iteration.