Skip to content
View jkanalakis's full-sized avatar

Highlights

  • Pro

Block or report jkanalakis

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
jkanalakis/README.md

πŸ’‘ About Me

🌟 Engineering Leader and AI Researcher with extensive experience in AI/ML, Streaming Video, and Full-Stack Development. I specialize in building innovative, scalable, and impactful products enjoyed by millions worldwide.
πŸ”’ Experienced at delivering secure, cutting-edge solutions in AI-powered systems, Web platforms, and Mobile applications.
πŸš€ Skilled at leading cross-functional teams, mentoring talent, and driving product success under tight timelines.
🀝 Dedicated to combining AI's intelligence with transparency and ethical practices to solve complex challenges.

πŸ“ˆ Career Highlights

  • Founded Tendi, an AI-powered financial advisor platform praised for its role in bridging the wealth gap.
  • Scaled Roku's international streaming platform to 6M+ active devices and optimized AI-driven content recommendations.
  • Designed and launched Netflix's first video streaming UI for Xbox 360, spearheading the transition to streaming services.
  • Designed Cisco's complete user interface framework for umi home video conferencing.
  • Advanced TiVo's award-winning embedded televisiona television user experience.
  • Pioneered advanced AI algorithms in personal finance, including RAG, RLHF, and fine-tuning for GPT-4, Llama3, and Mistral.
  • Directed high-performance engineering teams to deliver innovative solutions in video conferencing, streaming, and consumer electronics.

πŸ› οΈ My Tech Stack

Frontend Development

  • Frameworks & Libraries: SwiftUI, React.js, Vue.js, Next.js, Angular.js, Tailwind CSS, Gradio
  • Design Tools: Figma, Sketch, Lottie.js

Backend Development

  • Languages & Frameworks: Node.js, Django, FastAPI, Express.js, Nest.js, hapi.dev

Database Management

  • Technologies: MongoDB, PostgreSQL, MySQL, Supabase, Firebase

AI & Machine Learning

  • Models: OpenAI GPT, Meta Llama, Mistral, Google Vertex AI
  • Tools: PyTorch, TensorFlow, LangChain, LlamaIndex, CrewAI, Hugging Face, Scikit-learn
  • Techniques: RLHF, fine-tuning, data labeling, and vector indexing

DevOps

  • Platforms: AWS, DigitalOcean, Docker, Kubernetes
  • CI/CD: GitHub Actions, Jenkins

🌟 Key Skills

  • AI/ML: Model training, fine-tuning, and MLOps deployment
  • Leadership: Building and mentoring high-performance teams
  • Product Development: UX design, A/B testing, and user-centric solutions
  • Scalable Solutions: Microservices architecture and RESTful APIs

πŸš€ Current Projects

Tendi.ai

(Visit Website)

  • Description: Democratizing financial literacy with an AI-powered advisor.
  • Achievements:
    • Developed a personalized budgeting tool using advanced ML algorithms, a fine-tuned model, and extensive prompt engineering.
    • Leveraged NLP to understand users’ financial personas and social interests.
    • Launched a mobile app featured in CNET Money.

Finetuning LLaMA Model for Text Generation Using Unsloth

(Visit Repository)

  • Description: A research-driven project focused on fine-tuning Meta's LLaMA model to enhance its text generation capabilities using Unsloth, a novel technique for efficient data labeling and augmentation. The project optimizes the model's performance on domain-specific tasks by leveraging Unsloth's semi-supervised learning pipeline, enabling rapid iteration with minimal labeled data. Advanced techniques like LoRA (Low-Rank Adaptation) are used to fine-tune the model efficiently without requiring extensive computational resources.

Diabetic FoodPilot

  • Description: An AI-powered solution designed for diabetics to estimate carbohydrate consumption and plan insulin dosage effectively. The project leverages data from FoodCentral, location-based insights, and user-specific food preferences to calculate carb content and recommend optimal insulin dosages. It also considers the impacts of varying insulin dosages on glucose levels and lifestyle activities.
  • Tech Stack: LLMs for NLP-based food description parsing, PyTorch for AI model development, TensorFlow for prediction modeling, geolocation APIs for contextual data, and FoodCentral API for comprehensive nutritional data.

πŸ“š Past Experiences

  • Founder, Crediture (2021-2024): Created an AI-powered fintech startup to democratize small business lending.
  • Director of Engineering, Roku (2014-2021): Scaled international streaming business with AI-driven content recommendations.
  • Senior Engineering Manager, Netflix (2011-2014): Delivered innovative user interfaces and transitioned millions to streaming.
  • Senior Engineering Manager, Cisco (2009-2011): Assembled a team and developed the full user experience for in-home telepresence video conferencing.
  • Engineering Manager, TiVo (2006-2008): Led embedded Linux user interface development to pause, schedule, and record live broadcast television.

πŸ† Patents & Publications

  • Patents: Developed a TV-centric UI framework for Cisco’s Ε«mi telepresence product.
  • Publications:

πŸ“« Contact Me

πŸ’¬ Let's Collaborate!

I'm always interested in connecting with professionals passionate about AI, technology, and innovation. Whether you're exploring new ideas, looking for a mentor, or want to discuss cutting-edge tech, feel free to reach out!

Pinned Loading

  1. finetuning-llama-model-for-text-generation-using-unsloth finetuning-llama-model-for-text-generation-using-unsloth Public

    Fine-tuning Llama 3.2 3B Instruct model for text generation using Unsloth AI

    Jupyter Notebook 2