Matt Poetter

About Matt Poetter

Nearly 20 years of AI experience (Artificial Intelligence / Computer Vision):Key Skills: Deep Learning (CNN, RNN, TensorFlow, PyTorch, etc.), Natural Language Processing: LLMs, LMMs, RAG, LoRA, LlamaIndex, RAGChain, Parameter-Efficient Fine-Tuning Methods (PEFT) for Pretrained Language Models (PLMs), DPO, PPO, Graph Neural Networks (GNNs), Knowledge Graphs (KGs), LLMOps, LMMOps, RLHF, RLAIF, SSMs, H3, Mixture of Experts (MoE), MetaGPT, Reinforcement Learning (RL), Q* Search/Q-Learning), Policy NN, Value NN, MCTS, OpenAI CLIP Architecture, DocLLM.Data Science (Apache Spark MLlib, Mahout, R, spaCy, Anaconda), Hybrid Models (predefined structures + neural networks + weights/stochastics, e.g.), Hybrid Models (predefined structures + neural networks + weights/stochastics) e.g. LSTM (Long Short-Term Memory), GRU (Gated Recurrent Units), Attention, Feast AI), ONNX, PMML, OpenScoring.io, storage of deep learning intermediate results + models, knowledge representation and inference (reasoning, drawing conclusions), semantics, virtualisation, management with Docker, Kubernetes, Airflow, etc.

Experience

1. Artificial Intelligence Consultant

   Compris Technologies Corp

   Mar 2020 — Present · San Jose, CA, US

   Making startups and established companies become more competitive with latest AI techniques, especially intelligent Robotic Process Automation (RPA) harnessing latest deep learning insights for Natural Language Processing (NLP) and Computer Vision (CV) combined with other AI techniques.Purpose: Automation of activities- For administrative activities: Reading, editing, extracting and otherwise entering information into IT systems: In files, APIs,(web) forms, UIs of apps, etc.)- For agile processes: Automate as much as possible and identify and recommend situations for the application of now over agile practices, especially in risk and conflict management, seizing opportunities, identifying decision points, etc- For eLearning: Automated suggestion of tips, corrections, transfer of data, introduction of psychological elements for motivation, etc, among other things by injecting JavaScript (e.g. as an avatar), which analyses, gives tips, asks control questions about the learning content, etc- For competitive and market analysis (customers, suppliers, competitors, financial and personnel market): web-mining, web-scraping, information processing.

Education

• Stanford University

  Master of Computer Applications - MCA

  2007 — 2009

• Stanford University

  Bachelor's degree

  2003 — 2007