Adam Gabriel Email & Phone Number

New York, Ny, Us

New York, Ny, Us

New York City Metropolitan Area

Simplilearn is the world's #1 training bootcamp. I'm training cohorts (classes) at the Post-Graduate (PG) level in Artificial Intelligence & Machine Learning (AI/ML), as well as Advanced Generative AI. The practical, hands-on aspects of the course like the exercises, section-end projects, and final Capstone project are all coded in Python and its supporting AI/ML libraries.The curricula come from esteemed institutions like Purdue University and CalTech, and include courses from industry titans like IBM.

Tempe, Arizona, United States

Working on a variety of specialized projects, most recently and notably are Webmetheus, a web scraping tool, and ART, an Automatic Reporting Tool. For the backend we used:- Node.js- Node-REDFor the frontend:- JavaScript- Vue- Petite-Vue- Bootstrap (with some HTML and CSS tweaks, of course)Finally, each of those those tools, Webmetheus and ART, was converted from just having a web interface, into a custom Node-RED node, so that node could be incorporated into Node-RED flows that would run at different preset times for each of the company's clients, times that were read from a web-based configuration utility.

New York

To fine-tune a GPT (Generative Pre-trained Transformer) like ChatGPT for your specific use case, I typically use:1. LoRA (Low-Rank Adaptation): This reduces computational & memory needs when fine-tuning models like the Transformer that powers ChatGPT & other Large Language Models (LLMs). 2. QLoRA (Quantized Low-Rank Adaptation): Builds on LoRA by quantizing model weights, further optimizing memory & computational costs. Especially effective for large models like Meta’s LLaMA(2).OpenAI enabled LoRA for ChatGPT in August 20, 2024, & I jumped right on it. With my experience fine-tuning models like LLaMA 2, I can fine-tune ChatGPT-4o —the most powerful GPT— to deliver incredible precision for your domain.But here's the kicker: fine-tuned ChatGPT-4o still can NOT make API calls. That’s where I come in. I’ll create a custom GPT to handle API calls, automatically invoking actions for optimal responses. It’ll feel almost magical. 🪄When starting from scratch, Python is my go-to for building Neural Networks (NN). I prefer using Keras with a PyTorch backend (though I’m also proficient in TensorFlow). I’m skilled in NumPy, SciKit (sklearn), & Pandas to clean, normalize, & feature-scale data before feeding it to machine learning models. Matplotlib & PyPlot are my tools for visualization.Some NNs I’ve built:- A bank customer churn predictor based on customer data (sex, geolocation, income, # of accounts, etc.).- A CNN (Convolutional Neural Network) for image classification whose accuracy I boosted from 82% to 92% accuracy using dropout layers.- A RNN (Recurrent Neural Network) with LSTM (Long Sort Term Memory) layers to predict stock prices.- A movie recommendation system, an unsupervised learning AutoEncoder, using the usual 5⭐️ system. By tweaking the number of NN layers & neurons in each, I maximized prediction accuracy from 4 to 4.5 stars.Bottom line? I’ll fine-tune your GPT, create custom GPTs, or build powerful neural networks to meet your needs.

New York

Boy oh boy the stories I can tell here! Even though we had a huge inventory - $80 million worth of art, it was privately held, there were only a handful of employees, so it was both run as, and felt like, a startup.Lots of (almost) sleepless nights, blending into days of curated opening events with champagne and caviar. Wanna talk about it over coffee, or given the subject matter, shall we go for cheese and wine? ;-) lolWell, maybe I'll share some of my tales. Watch this space. Stay tuned.