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 ↗
LlamaIndex Agentic RAG and Workflow Engine
Jerry Liu · LlamaIndexModular data framework connecting unstructured enterprise knowledge bases with tool-calling agents through automated query routing and sub-question synthesis.
Retrieval accuracy depends on optimal chunking strategies and embedding domain specificity.
Context: Python/TypeScript, vector databases, and LLM query transformation.
View missionDeep Medicine: High-Performance AI in Clinical Diagnostics and Patient Care
Dr. Eric Topol · Scripps ResearchClinical evaluation assessing the diagnostic accuracy, speed, and safety boundaries of multimodal AI models across cardiology, radiology, and oncology.
Clinical diagnostic deployment requires FDA approval and institutional ethics review; model outputs must assist rather than replace licensed physician judgment.
Context: Multimodal clinical models, ECG neural nets, and electronic health record parsing.
View missionLarge-Scale Generative AI Red-Teaming Methodology
Dr. Rumman Chowdhury · Humane IntelligenceCommunity red-teaming protocols orchestrating thousands of independent researchers to evaluate bias, safety boundaries, and prompt exploits in commercial foundation models.
Evaluates public checkpoint behavior; closed proprietary internal corporate systems require continuous telemetry auditing.
Context: Adversarial prompt evaluation and bias detection metrics.
View missionPatterns for Building LLM-based Systems & Products
Eugene Yan · AmazonComprehensive reference manual documenting production patterns for evaluation, guardrails, retrieval-augmented generation, and fine-tuning with cost/latency trade-offs.
System design patterns; individual implementations must be profiled for specific latency constraints.
Context: Python, vector indexes, cross-encoders, and LLM evaluation frameworks.
View missionThe AI Engineer: The Rise of the Fastest Growing Role in Software Engineering
Swyx (Shawn Wang) · Smol.ai / Latent SpaceSeminal technical essay analyzing the shift from classical ML research to application-layer AI engineering, defining evaluation, prompt caching, reasoning scaffolding, and tool integration.
Framework assumes APIs and open weights continue advancing capabilities faster than traditional specialized training.
Context: Foundation model APIs, TypeScript, Python, and developer tooling.
View missionLLM-Powered Autonomous Agents: Comprehensive System Design & Architecture Analysis
Dr. Lilian Weng · Independent / ex-OpenAILandmark reference architecture decomposing autonomous agents into planning (subgoal decomposition, reflection), memory (short-term in-context, long-term vector search), and tool execution loops.
Identifies fundamental limits of finite context windows, reliability bottlenecks in long-term planning, and unreliability of open-ended natural language tool interfaces.
Context: Large language models, vector memory systems, ReAct prompting, and agent loops.
View missionAbu Dhabi Investment Authority (ADIA): Domain-Specific AI Evaluation Benchmark & Methodology
Dr. Marcos Lopez de Prado · Abu Dhabi Investment Authority (ADIA)An empirical evaluation methodology developed by Dr. Marcos Lopez de Prado at Abu Dhabi Investment Authority (ADIA), establishing standardized performance metrics, domain-specific evaluation sets, and error taxonomy protocols.
Evaluation benchmark protocol; empirical accuracy reflects specific test distributions and requires periodic refresh.
Context: Mathematical finance, combinatorial cross-validation, and portfolio optimization.
View missionpH-7 Open-Source Web Framework Architecture & AI-Integrated Software Suite
Pierre-Henry Soria · Soria / pH-7Public open-source software suite implementing modular OOP architecture, test-driven development (TDD), and extensible AI microservices with over a decade of production maintenance.
Requires modern PHP 8.2+ or Node.js runtime environments with relational database backends (PostgreSQL, MySQL).
Context: Open-source GitHub repository with automated CI/CD test suites, static analysis, and Docker configurations
View missionVoyager: An Open-Ended Embodied Agent with Large Language Models
Dr. Jim Fan · NVIDIALifelong learning autonomous agent architecture featuring an automated curriculum, self-expanding skill library of executable code, and iterative prompting mechanism with environment feedback.
Dependent on real-time vision-language reasoning latency; physical embodiment requires sim-to-real transfer validation.
Context: GPT-4, Mineflayer JavaScript bot API, vector database skill retriever, and Python.
View missionVoyager: An Open-Ended Embodied Agent with Large Language Models
Dr. Linxi (Jim) Fan · NVIDIAGroundbreaking research introducing Voyager, the first LLM-powered embodied lifelong learning agent in Minecraft that continuously explores world dynamics, acquires skills, and makes novel discoveries without human intervention.
Evaluated within Minecraft simulation environment; real-world robotic embodiment introduces hardware physical latencies and physical sensor noise.
Context: GPT-4 prompting, iterative prompting mechanism, and skill library retrieval.
View missionbitsandbytes: 8-Bit and 4-Bit CUDA Quantization Primitives
Tim Dettmers · Allen Institute for AI (Ai2)The open-source library behind 8-bit matrix multiplication and 4-bit NormalFloat quantization, enabling consumer GPUs to run and train 70B+ parameter foundation models with negligible accuracy loss.
Tightly coupled with NVIDIA CUDA architectures; requires compilation of custom C++/CUDA kernels for optimal memory coalescing.
Context: CUDA, C++, Python, Hugging Face transformers, PEFT, and PyTorch.
View missionBuilding LLM-Based Systems: Latency, Cost, and Architecture Trade-offs
Chip Huyen · Independent / Claypot AIComprehensive guide detailing the engineering differences between training, fine-tuning, and prompt-based RAG architectures in enterprise production.
System design blueprint; implementation performance depends on team infrastructure and latency budgets.
Context: LLM systems engineering, inference caching, and evaluation harnesses.
View missionBabyCatAGI: Parallel Task Execution and Priority Orchestration Engine
Yohei Nakajima · Untapped VCAn advanced iteration of the autonomous BabyAGI architecture created by Yohei Nakajima, introducing parallel agent threads, dynamic task prioritization, and structured context sharing.
Autonomous agent loops can recurse indefinitely without strict budget stop conditions; relies on foundation model task completion signals.
Context: Python, Chroma/Pinecone vector storage, and OpenAI API.
View missionGuardrails: Adding Guardrails to Large Language Models with Structured Validation
Shreya Rajpal · Guardrails AIAn open-source specification language (RAIL) and runtime engine enforcing structure, type, and semantic quality constraints on LLM outputs with automated corrective re-asking loops.
Corrective re-asking loops introduce additional inference latency and token overhead on malformed generations.
Context: Python, Pydantic, JSON Schema validation, and multi-model LLM APIs.
View missionPyHealth: A Deep Learning Toolkit for Healthcare AI and Clinical Predictive Modeling
Dr. Jimeng Sun · University of Illinois Urbana-Champaign (UIUC)Modular PyTorch toolkit standardizing clinical machine learning workflows across MIMIC, eICU, and OMOP datasets for mortality prediction, readmission risk, and drug recommendation.
Requires HIPAA-compliant data environments and de-identification pipelines when handling raw electronic health records.
Context: PyTorch, Python, OMOP Common Data Model, MIMIC-IV, and clinical embeddings.
View missionnanoGPT: Minimal, Readable, and Fast PyTorch Framework for Training Medium GPTs
Andrej Karpathy · Eureka Labs / ex-Tesla, OpenAIA pure, clean, and hackable 300-line PyTorch implementation for training, fine-tuning, and evaluating transformer language models with support for FlashAttention and torch.compile.
Engineered for educational transparency and small-to-medium parameter models; multi-node distributed trillion-token training requires complex pipeline parallel frameworks like Megatron.
Context: PyTorch 2.0, CUDA, Python, FlashAttention, and OpenWebText.
View mission