AI research and projects
The research, projects, and explanations behind the people. Read the source, inspect verified code and models, and find a collaborator.
A source check confirms that a source supports the summary. It is not an independent replication or endorsement. Our methodology ↗
Fourier Neural Operator for Parametric Partial Differential Equations (FNO)
Anima Anandkumar · California Institute of Technology (Caltech)Pioneered the Fourier Neural Operator (FNO), formulating a deep learning architecture that directly learns mappings between infinite-dimensional function spaces via Fast Fourier Transforms, achieving mesh-independent zero-shot super-resolution on Navier-Stokes fluid turbulence 1,000x faster than standard numerical solvers.
Periodic boundary condition assumptions require specialized padding/coordinate embeddings for complex arbitrary 3D boundaries.
Context: Fourier spectral convolution layers evaluated on 2D/3D Navier-Stokes, Darcy Flow, and global atmospheric weather forecasting (FourCastNet).
View missionFrom Scratch: Intuitive Vector Search & Embedding Spaces
Santiago Valdarrama · UnderfittedAuthored an interactive architectural breakdown explaining how high-dimensional vectors, dot-product projections, cosine similarities, and Hierarchical Navigable Small World (HNSW) graphs work under the hood.
HNSW approximate nearest neighbors algorithms trade off index build time and RAM consumption for query latency.
Context: Python, NumPy, FAISS, and Chroma vector database architectures.
View missionGemini: A Family of Highly Capable Multimodal Models
Jeff Dean · Google / Google DeepMindCo-led the executive and technical direction of Gemini, designed natively from the ground up to interleave and reason across text, image, video, audio, and code tokens across Google TPU v4 and TPU v5e supercomputing pods.
Interleaved audio-video processing requires aggressive token pruning to sustain real-time serving latency.
Context: Native multimodal transformer trained on TPU v4/v5e clusters with million-token context windows.
View missionGitHub Copilot: Conception, Technical Architecture & Launch
Nat Friedman · AI Grant / Andromeda ClusterConceived and executive-led GitHub Copilot, pairing OpenAI Codex with IDE telemetry, context window assembly from local tabs, and low-latency inline code ghost-text completions used by millions of software developers.
Early versions could suggest code containing outdated security vulnerabilities or license incompatibilities without automated scanning.
Context: OpenAI Codex model running custom token-level streaming inference into Visual Studio Code extensions.
View missionGloVe: Global Vectors for Word Representation
Christopher Manning · Stanford UniversityCreated the GloVe log-bilinear word embedding model that combines the advantages of global matrix factorization and local context window methods, capturing linear substructures and semantic relationships in vector space.
Static word vectors conflate multiple distinct polysemous meanings of words into a single point representation.
Context: Trained on Common Crawl 42B and 840B token corpora.
View missionGoogle Data Center Cooling Optimization via Deep Neural Networks
Mustafa Suleyman · Microsoft AILed the applied DeepMind engineering team that deployed deep neural networks to autonomously manage cooling infrastructure across Google's massive global data centers, slashing cooling energy consumption by 40%.
Requires human-in-the-loop safety boundaries to override automated actuator commands during unexpected weather spikes.
Context: Ensemble of deep neural networks processing historical sensor streams (temperatures, pressures, power) to predict future energy efficacy.
View missionGPT-4 System Developer Keynote & Real-Time Multimodal Demonstrations
Greg Brockman · OpenAIConducted the historic live technical demonstration of GPT-4, parsing a handwritten website mockup sketched on a napkin into functional HTML/JavaScript code in real time, and debugging complex tax calculations across a 16-page IRS document.
Handwritten input legibility and complex nested mathematical tables require multiple-pass validation.
Context: GPT-4 multimodal vision checkpoint with 32k context window and zero-shot visual reasoning.
View missionGPT-4o: End-to-End Real-Time Multimodal Omni Model Architecture
Mira Murati · Independent AI Research LabLed the research and product execution of GPT-4o, unifying text, visual, and audio tokens into a single native neural network that achieves 232ms average voice response latency, matching human conversational response intervals while preserving complex reasoning.
Real-time speech interruption handling requires aggressive client-side endpointing and noise cancellation to avoid cutting off human speakers.
Context: Native end-to-end multimodal transformer trained on text, audio, and visual tokens simultaneously.
View missionGrandmaster Level in StarCraft II using Multi-Agent Reinforcement Learning (AlphaStar)
Oriol Vinyals · Google DeepMindLed the AlphaStar research team combining imitation learning, multi-agent league training, and deep transformer-based recurrent policies to master StarCraft II—operating under imperfect information, long-horizon real-time strategy, and high-dimensional action spaces.
League diversity must be actively curated with dedicated exploiters to prevent cyclic policy stagnation.
Context: Multi-agent league self-play across 35 days of TPU v3 training per agent with human-like action rate limits.
View missionHands-on Fine-Tuning Open Source LLMs for Production Tasks
Santiago Valdarrama · UnderfittedCreated open-source walkthroughs showing practitioners how to prepare structured JSONL datasets, apply LoRA adapters via Hugging Face PEFT, and quantize models using bitsandbytes on a single GPU.
Single-GPU fine-tuning restricts effective batch sizes without gradient accumulation and activation checkpointing.
Context: Llama 3 8B, Mistral 7B, PyTorch, Hugging Face TRL and PEFT.
View missionImageNet Classification with Deep Convolutional Neural Networks (AlexNet)
Geoffrey Hinton · University of TorontoCo-authored the milestone AlexNet paper that crushed the ImageNet competition in 2012 by an unprecedented 10.8% margin, demonstrating the power of GPU-accelerated convolutional networks with ReLU activations and Dropout regularization.
Required heavy data augmentation (cropping, flipping, PCA jittering) to prevent severe overfitting on large parameter weights.
Context: 8-layer CNN trained across two NVIDIA GeForce GTX 580 GPUs on 1.2 million high-resolution images.
View missionImproving Language Understanding by Generative Pre-Training (GPT-1)
Ilya Sutskever · Safe Superintelligence Inc. (SSI)Co-architected the foundational framework establishing that unsupervised autoregressive pre-training on diverse text corpora, followed by discriminative task fine-tuning, produces superior transfer learning across reasoning, classification, and natural language inference benchmarks.
Fine-tuning still required bespoke task heads prior to the zero-shot prompting breakthrough of GPT-2 and GPT-3.
Context: 12-layer decoder-only transformer with masked self-attention heads trained on BooksCorpus.
View missionInflection-1 & Pi: Empathetic High-Fidelity Conversational Agents
Mustafa Suleyman · Microsoft AISpearheaded the training and deployment of Inflection-1 and Inflection-2, powering the Pi conversational assistant with natural voice synthesis, low-latency conversational turn-taking, and empathetic persona tuning.
Empathetic conversational tone requires strict guardrails to prevent emotional over-dependence in vulnerable users.
Context: Proprietary foundation model trained on a 22,000 NVIDIA H100 GPU cluster with FP8 mixed precision.
View missionLearning Representations by Back-Propagating Errors
Geoffrey Hinton · University of TorontoAuthored the foundational 1986 Nature paper demonstrating that backward propagation of error gradients enables multi-layer neural networks to learn internal representations, resolving the perceptron limits and founding modern connectionist AI.
Susceptible to vanishing/exploding gradients in extremely deep architectures prior to residual connections and normalized activations.
Context: Multi-layer perceptrons with generalized delta rule gradient descent.
View missionMastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm (AlphaZero)
David Silver · Google DeepMind / UCLCreated AlphaZero, which mastered Go, chess, and shogi starting completely from scratch tabula rasa given only the game rules, crushing world-champion engines Stockfish and Elmo within 24 hours of self-play.
Assumes deterministic, fully observable game dynamics with explicit simulator rules.
Context: Single deep residual neural network updating continuous MCTS action evaluations across 5,000 TPU v1s.
View missionMastering the Game of Go with Deep Neural Networks and Tree Search (AlphaGo)
David Silver · Google DeepMind / UCLLed the AlphaGo project combining deep value networks and policy networks with Monte Carlo Tree Search (MCTS), solving the 2,500-year-old game of Go and defeating 18-time world champion Lee Sedol 4-1 in Seoul.
Initial AlphaGo version relied on 30 million human expert games for supervised pretraining prior to self-play.
Context: 12-layer deep convolutional policy and value networks trained across 48 Google TPUs.
View mission