I'm Jack Felke— AI Engineer at All Digital Rewards. I build innovative, commercial-grade projects and open source tools that push the envelope in AI and full-stack development. I'm not just an enthusiast; I work professionally on solutions that scale.

I believe in continuous improvement both personally and professionally. Inspired by the habit stacking method from Atomic Habits by James Clear, I design bulletproof routines that include breathwork, cardio, and walks with Stella. I'm proud of my current 42-day streak on Duolingo as I work on mastering both Chinese and French.

In 2025, I'm ready to launch several new projects and tackle fresh challenges at work. I’m excited about evolving my skills in modern web frameworks, containerization, and AI engineering while maintaining a healthy, balanced lifestyle.

• Boost Productivity: Use habit stacking to create routines incorporating breathwork, cardio, and walks for focus and energy.
• Learn Languages: Maintain my Duolingo streak to master Chinese and French.
• Live Sustainably: Build a sustainable routine that supports creativity and well-being.

• Launch Projects:
  • Hearthful App: An AI-powered event planning tool for large groups and family gatherings.
  • Diagram Canvas Project: A Next.js-based tool for generating and editing diagrams using Excalidraw JSON and Mermaid.js.
    • Built with LangGraph for agent workflows, shared states, and nodes for logic.
    • Features include diagram previews, natural language edits, image uploads, code copying, and PNG downloads.
  • AI Part Picker
  • Prompt Engineering App: Web app + VS Code extension for on-device codebase vector stores, cloud storage, and context-aware prompts.
  • Realtime API Tools: Chrome and VS Code extensions for voice-assisted coding.
  • Open Source DevTools: CLI tools and developer utilities.
  • YouTube AI Extension
• On-Device LLMs: Experiment with fine-tuned LLMs on Android using small models like Phi-3b.
• iOS Development: Learn UIKit and build apps for iOS, watchOS, macOS, and visionOS.
• AI Closet Stylist: An AI that learns your wardrobe and suggests outfits.

• Master SSR Frameworks: Deepen expertise in Angular (with Material) and Node.js.
• Advance Containerization: Continue working with Docker.
• Improve AI Pipelines: Enhance RAG pipelines with evaluations, callback tracing, guardrails, agentic improvements, and model tuning.
• Master LangGraph: Fully understand LangGraph by Q1 2025 to optimize AI workflows.

I’m also tracking advancements in DeepSeek R1, Kimi k1.5, optimization methods, and RL frameworks. Exploring model distillation to fine-tune smaller models using high-quality data from larger models, leveraging web data, vector stores, and reasoning tokens for better performance.

• Languages & Frameworks:

  • 
  • 
  • 
  • 
  • 

• UI Component Libraries (Next/Vite/Vue):

  • 
  • 

• UI Design AI Assistants:

  • 
  • 

• State Management & Libraries:

  • 

• Languages & Frameworks:
  • 
  • 
  • 

• 

• 

• Techniques & Methodologies:

  • I specialize in:
    • 
    • 
    • 
    • 

• Frameworks & Libraries:

  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 
  • 

• Databases & Hosting:
  • 
  • 
• Cloud & DevOps:
  • 
  • 
  • 
  • 

• Tools:
  • 
  • 
  • 
  • 
  • 
  • 

• Tools:
  • 
  • 

I'm working on a clone of ChatGPT Canvas that generates and renders diagrams using Excalidraw JSON and Mermaid.js.

• Tech Highlights:
  • I built this as a Next.js app powered by LangGraph, which models agent workflows as graphs.
  • I designed a workflow where:
    • States are shared data structures (Annotation objects) that capture the current snapshot.
    • Edges are functions that determine which node to execute next based on the current state.
    • Nodes encapsulate the logic for both the o3-mini and my fine-tuned 4o model agents.
  • The o3-mini works alongside the 4o model to refine the generated code. With this project, I can preview diagrams, edit them using natural language instructions, upload images, copy code, and download diagrams as PNG.

Note: I'll update this section with the latest details on finetuning, dataset generation, and workflow improvements.

I'm developing an AI-powered event planning tool for large groups or family gatherings. I started prototyping this project in December, and I can't wait to see it come to life.

I have half a dozen projects planned for 2025, blending commercial use cases with open source contributions that leverage my expertise in full-stack development and AI engineering.

At All Digital Rewards, I've experienced tremendous growth:

• Key Achievements:
  • I deployed production code for an AI chatbot featuring role-based context access and hundreds of vector stores.
  • I built scalable solutions using TypeScript, Bun.js, Docker, TypeORM, Qdrant vector stores, and LangChain concepts.
• Learning & Growth:
  • I embraced new challenges by learning Angular (with Material) and exploring various SSR frameworks.
  • I expanded my skills in containerization with Docker and am excited to soon learn Kubernetes.
  • I continuously work on improving RAG pipelines through evaluations, callback tracing, guardrails, agentic improvements, embedding optimizations, and model tuning/distillation.
• My Philosophy:
  I believe coding is like working out—the more consistently I practice, the stronger I become. I know that small wins add up, and I strive to maintain balance and momentum.

• Personal Website: terminalgravity.dev (Coming Soon)
• LinkedIn: My LinkedIn
• X: @JackFelke
• Email: [email protected]