(Fullstack) AI Engineer
All engineers should be able to build AI native products across all platforms (web, mobile, desktop) at high velocity and independently.
I am still seeing engineering teams composed of backend, frontend AND Machine Learning engineers. Such separation of concern fractures the potential user-experience AI native apps can provide. Web browsers have become insanely powerful and many new browser API like Server-side Events unlock net new capabilities that can help build sleek UX. Models will keep improving and teams will only try to fine-tune them once the usecase is well understood, evals are well-defined for said usecase and the team is looking to optimize for
AI Engineering has more concerns in the application layer
Below is the minimum bar for engineers on my teams with resources to help them upskill and fill gaps. This bar is the floor, not the ceiling. I don’t expect everyone to be above the floor on day one. All new hires must meet this bar to be eligible for interviews.
Behavioral¶
- Ownership: Take end-to-end responsibility for tasks or projects. Take initiative to solve problems that you see.
- Curiosity & Continuous Learning: Stay on top of new AI frameworks and best practices.
- Collaborative Problem-Solving: Work openly and effectively with other team members.
- High-Velocity Mindset: Optimize for quick iteration, shipping MVPs, and learning.
- Show-and-Tell: Talk is cheap, and prototypes are cheaper. Show instead of making proposals.
- User obsessed: Design technology that makes people’s experience more natural and seamless. We care about building an exceptional user experience.
- Detail-focused: Exceptional outcomes come from exceptional attention to detail to every interaction and every pixel.
Technical¶
| – | – |
|---|---|
| Languages | Typescript, SQL, Bash, HTML5, CSS3, Python |
| Frameworks/Runtime | React, NextJS, React Query, Zustand, React Native, Expo, TailwindCSS, RadixUI, DrizzleORM, Zod, NodeJS, WebRTC, PyTorch |
| Tooling | Turborepo, ESLint, Prettier, NPM/PNPM, Git, AI codegen |
| Databases | Postgres (incl PGVector, JSON datatype), Object stores |
| Infrastructure | Docker, Cloudflare (Workers, Durable Objects, Pages, DNS, CDN), Vercel, Inngest, Render, Trigger.dev, Together.ai, Hugging Face, Replicate, OpenTelemetry (concepts) |
| AI Engineering | Large Language Models, Multi-Modal Models, Image Generation, In-painting, Voice Models |
| AI Concepts (learn over time) | Basic Prompting, Chain of Thought prompting, N-shot prompting, Prompting reasoning models, Tool calling, Preprocessing unstructured data, ReAct, Agent basics, Advanced agentic patterns, Evals, Memory, Generative UI, Streaming, Real-time, Multi-agent systems, Guardrails, Citations, Vector databases, RAG, Text embeddings, Knowledge graphs, Query routing, Synthetic data, Fine-tuning (SFT, RL, DPO, GRPO), Diffusion Models, RLHF, MCP, Computer use, Using/Serving Multi-modal OSS models |