PUBLISHED AI RESEARCH AND PROJECTS

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 ↗

Showing 17–32 of 567 proofs of work · Newest first
Page 2 of 36
implementationChecked Sep 25, 2026

Salesforce Agentforce: Autonomous Enterprise Multi-Agent Orchestration

Clara Shih · Meta / ex-Salesforce

Spearheaded the product architecture of Agentforce, enabling enterprises to deploy autonomous AI agents with unified CRM data grounding, Atlas reasoning engines, and policy guardrails that operate without rigid hardcoded decision trees.

Scope & limitations

Enterprise agent reliability requires high data cleanliness and granular field-level permissioning across fragmented legacy enterprise systems.

Context: Atlas Reasoning Engine, Einstein Trust Layer, CRM Data Cloud grounding.

View mission
implementationChecked Sep 25, 2026

Snorkel: Rapid Training Data Creation with Weak Supervision

Alex Ratner · Snorkel AI / University of Washington

Created the Snorkel open-source system and paradigm, using generative probabilistic modeling to combine noisy heuristic rules, knowledge bases, and small models into high-accuracy probabilistic training labels without manual annotation.

Scope & limitations

Heuristic labeling functions can introduce correlated bias if domain experts write overlapping or conditionally dependent heuristics without proper graph independence modeling.

Context: Probabilistic generative label models, matrix completion, weak supervision algorithms.

View mission
explanationChecked Sep 25, 2026

The AI 3.0 Stack: Investment Thesis on Autonomous Native Systems

Sarah Guo · Conviction / No Priors

Authored the foundational AI 3.0 thesis outlining the transition from superficial LLM wrappers to vertically integrated, feedback-loop-driven AI systems with proprietary eval datasets and cognitive agent workflows.

Scope & limitations

High compute costs and foundation model commodity pressure can compress margins for companies lacking proprietary workflow lock-in.

Context: AI infrastructure, proprietary evaluation loops, continuous fine-tuning, domain-specific vertical agents.

View mission
explanationChecked Sep 25, 2026

The Blueprint for Enterprise Autonomous Agents in Customer Workflows

Clara Shih · Meta / ex-Salesforce

Formulated the enterprise framework contrasting predictive machine learning, conversational copilot bots, and fully autonomous action-taking agents capable of resolving multi-step business cases end-to-end.

Scope & limitations

Organizational cultural resistance and legacy service metrics (such as handle time versus resolution quality) complicate enterprise agent deployment.

Context: Enterprise CRM workflows, human-in-the-loop escalation matrices, generative customer interactions.

View mission
explanationChecked Sep 25, 2026

The Shift from Monolithic Models to Compound AI Systems

Ali Ghodsi · Databricks

Co-authored the influential technical treatise establishing why state-of-the-art AI applications are moving from single monolithic LLMs to compound AI systems integrating multi-step reasoning, vector databases, code execution, and external verification tools.

Scope & limitations

Multi-component compound architectures multiply system failure modes, requiring end-to-end tracing and specialized evaluation harnesses.

Context: Compound AI architectures, RAG pipelines, LLM routing, Apache Spark, MLflow.

View mission
implementationChecked Sep 25, 2026

Together Inference Engine & FlashAttention-Optimized Model Serving

Vipul Ved Prakash · Together AI

Engineered Together AI's custom inference kernel stack, implementing speculative decoding, custom FP8 FlashAttention kernels, and continuous batching to deliver the industry's fastest serving latency for Llama 3, DeepSeek, and Mixtral.

Scope & limitations

Highly optimized custom kernels require aggressive quantization and continuous hardware profiling for non-standard GPU architectures.

Context: Llama 3/3.1, Mixtral 8x22B, custom CUDA/Triton kernels, speculative sampling, continuous batching.

View mission
explanationChecked Sep 25, 2026

Why AI Agents Represent the Next Fundamental Computing Platform

Bret Taylor · Sierra / OpenAI

Detailed the macroeconomic shift from static graphical user interfaces to autonomous conversational enterprise agents capable of goal-directed reasoning, tool invocation, and bidirectional system of record reconciliation.

Scope & limitations

Enterprise agent adoption velocity is gated by legacy API latency, messy unstructured enterprise data, and compliance audits.

Context: Enterprise software economics, tool-calling agents, human-in-the-loop validation patterns.

View mission
explanationChecked Sep 25, 2026

Why Search is the Hardest and Most Critical Foundation for Enterprise Generative AI

Arvind Jain · Glean

Authored technical analysis demonstrating that generative LLM outputs in corporate settings are only as accurate as the underlying retrieval precision, detailing why semantic vector search alone fails without enterprise graph metadata.

Scope & limitations

Synthesized enterprise answers risk exposing sensitive internal corporate gossip or draft documents unless strict contextual access filtering is applied.

Context: Vector retrieval, semantic search, precision@k metrics, context window stuffing vs targeted synthesis.

View mission
researchChecked Sep 24, 2026

A Style-Based Generator Architecture for Generative Adversarial Networks (StyleGAN)

Tero Karras · NVIDIA Research

Invented the style-based generator architecture for GANs, introducing adaptive instance normalization (AdaIN) to separate high-level facial attributes (pose, identity) from stochastic variations (hair, freckles).

Scope & limitations

Susceptible to characteristic water-droplet artifacts in feature maps before the StyleGAN2 weight demodulation redesign.

Context: FFHQ dataset, NVIDIA DGX-1 clusters, custom CUDA kernels.

View mission
implementationChecked Sep 24, 2026

Agent.ai: Multi-Agent Professional Workflow Directory & Execution Engine

Dharmesh Shah · HubSpot / Agent.ai

Architected a global discovery and execution platform for specialized AI agents that orchestrate multi-step research, prospect intelligence, competitive synthesis, and automated business workflows.

Scope & limitations

Autonomous agent execution requires strict human-in-the-loop review for state-changing enterprise actions to prevent cascading hallucinated writes.

Context: Distributed agentic orchestrator, LLM tool-calling APIs, dynamic web sandboxes.

View mission
implementationChecked Sep 24, 2026

ChatSpot.ai: Natural Language Conversational CRM Layer

Dharmesh Shah · HubSpot / Agent.ai

Developed one of the first production enterprise generative AI integrations combining OpenAI GPT-4 with structured enterprise CRM relational databases, executing natural language SQL queries and automated prospecting.

Scope & limitations

Schema complexity across enterprise custom objects necessitates strict few-shot prompt guards to avoid incorrect relational joins.

Context: GPT-4, HubSpot API, semantic intent parsers, relational data synthesizers.

View mission
implementationChecked Sep 24, 2026

Diffusers: State-of-the-Art Diffusion Models in PyTorch

Suraj Patil · Hugging Face

Engineered core schedulers (DDIM, DPMSolver, Euler), memory-efficient cross-attention kernels, and quantized pipeline architectures powering Stable Diffusion and Flux ecosystem inference.

Scope & limitations

Diffusion sampling requires multiple sequential denoising steps, bottlenecking latency relative to single-pass adversarial models.

Context: PyTorch, CUDA, xFormers, FlashAttention, ONNX runtime.

View mission
implementationChecked Sep 24, 2026

Diffusion Model Distillation & Latent Consistency Optimization

Sayak Paul · Hugging Face

Co-authored and integrated Latent Consistency Model (LCM) LoRA adapters into Hugging Face Diffusers, compressing multi-step diffusion sampling into 2 to 4 inference steps for real-time generation.

Scope & limitations

Aggressive 2-step distillation can introduce subtle high-frequency smoothing in photorealistic background details.

Context: PyTorch, CUDA, LCM-LoRA, SDXL, Apple Silicon Metal.

View mission
researchChecked Sep 24, 2026

Emergent Social Learning via Multi-Agent Reinforcement Learning

Natasha Jaques · University of Washington / Google DeepMind

Demonstrated that multi-agent reinforcement learning with intrinsic social curiosity incentives enables agents to spontaneously acquire social learning, imitation, and cultural transmission without human expert demonstrations.

Scope & limitations

Social imitation policies can converge to suboptimal collective traps if demonstrator agents experience local reward plateaus.

Context: Deep RL, multi-agent gridworld environments, social reward functions.

View mission
explanationChecked Sep 24, 2026

Fine-Tuning Stable Diffusion with LoRA and ControlNet

Suraj Patil · Hugging Face

Authored the definitive implementation and engineering guide for Low-Rank Adaptation (LoRA) applied to cross-attention layers of text-to-image latent diffusion models, slashing VRAM training requirements from 80GB to under 8GB.

Scope & limitations

Low-rank approximations can degrade fine textural fidelity when fine-tuning across drastically different visual domains.

Context: Parameter-efficient fine-tuning (PEFT), cross-attention adaptation.

View mission
implementationChecked Sep 24, 2026

GitHub Copilot: Real-Time Inline Generative Code Completion Architecture

Alex Graveley · Minion AI / Former GitHub

Conceived, architected, and built the initial GitHub Copilot production system, engineering client-side AST context window chunking, token speculative decoding, and low-latency streaming to VS Code.

Scope & limitations

Context window constraints in early versions restricted completions to single-file buffers before repository-wide semantic indexers were introduced.

Context: OpenAI Codex, TypeScript, VS Code language client, telemetry evaluation loops.

View mission
Showing 17–32 of 567 proofs of work