Loading use case index…
Loading use case index…
Daily updates on real-worldAI deployments worldwide.
Daily updates on real-world AI deployments worldwide.
Showing 722 results · 610 organizations · 62 industries · 57 countries/regions · sorted by Updated
Dartmouth's NILE Laboratory at the Geisel School of Medicine operates AI Patient Actor 2.0, an open-access platform where health-professions trainees interview simulated patients by text or emotive voice in 52 languages. A mixture of LLMs plays the patient, drafts cases and scores encounters, but every response is constrained by educator-authored Patient Profiles, Configurations and faculty rubrics. Students get rubric-scored feedback after each encounter, with prosody analysis for voice sessions. | 14 Nov 2026 | Dartmouth College | Diversified Consumer Services | United States |
SBI Securities deployed RightTouch's QANT Speak AI operator at the phone entry point to its customer centre, using short follow-up questions to identify the inquiry and route the call, and integrated it with Genesys Cloud so the AI's summary appears on the agent's incoming-call screen. After about a month of refining inquiry categories, tags and flows, routing accuracy exceeded 90%; average call time fell by 1.6 minutes, and some inquiry reasons reached 85% self-resolution. | 13 Nov 2026 | SBI Securities Co., Ltd. | Capital Markets | Japan |
Georgia Tech's Partnership for an Advanced Computing Environment (PACE) built SPIN (Service for Persistent INference), a fully on-premises AI gateway that needs no on-premises cloud. LiteLLM and PostgreSQL run as a Podman pod managed by systemd Quadlet on a CPU-only VM, while vLLM and Triton Inference Server run under Apptainer as Slurm jobs that scrontab keeps resubmitting and that re-register their host, port and key with LiteLLM. It is in limited availability for pilot groups of students and researchers. | 13 Nov 2026 | Georgia Institute of Technology | Diversified Consumer Services | United States |
Verge Genomics' CONVERGE 1.0 platform picked PIKfyve as an ALS target from a multi-omic analysis of nearly 1,000 patient and control tissue samples, and its PIKfyve gene network supplied GPNMB as the biomarker used to confirm target engagement and set dosing. The resulting drug, VRG50635, missed its efficacy endpoint in a 54-patient Phase 1b, exposing gaps in patient heterogeneity and blood-to-brain inference that Verge is addressing with generative models in CONVERGE 2.0. | 12 Nov 2026 | Verge Genomics | Biotechnology | United States |
The University of Münster has operated uniGPT, an on-premises LLM platform on its existing OpenStack-based Kubernetes cluster, since May 2024. Over eight design iterations it added a LiteLLM gateway for an OpenAI-compatible API with keys and quotas, moved inference from Hugging Face TGI to vLLM, and added image generation, speech recognition and RAG services. Over 99% of usage now comes through the API rather than the chat frontend. | 12 Nov 2026 | University of Münster | Diversified Consumer Services | Germany |
Hawaii Pacific Health Medical Group piloted Abridge integrated with Epic across 23 specialties. Among high users, time in notes per appointment fell from 7.14 to 5.63 minutes, a 21% reduction, while the study found no significant improvement in burnout. | 11 Nov 2026 | Hawaii Pacific Health Medical Group | Health Care Providers & Services | United States |
Peking University Cancer Hospital integrated a traceable AI documentation system with its EMR and clinical data centre. It reduced average record-writing time from about 3 minutes to under 20 seconds, using multimodal models and source-linked clinical facts. | 11 Nov 2026 | Peking University Cancer Hospital | Health Care Providers & Services | China |
The University of Idaho has run MindRouter, an open-source LLM inference gateway, in production since early 2026 on 13 GPU nodes with 40 GPUs serving 64 open-weight models across Ollama and vLLM backends. The FastAPI gateway translates among OpenAI, Ollama and Anthropic client protocols through a canonical intermediate schema and orders requests with a Weighted Deficit Round Robin fair-share scheduler. Azure AD single sign-on through Microsoft Entra ID maps institutional groups to scheduler weights and quotas, with audit logging and GPU telemetry sidecars keeping sensitive workloads on campus. | 11 Nov 2026 | University of Idaho | Diversified Consumer Services | United States |
Arizona State University launched Meletai in fall 2026 at its John Shufeldt School of Medicine and Medical Engineering. A multi-agent Case Creator workflow on ASU's CreateAI platform builds each patient case under faculty review; students interview a video avatar generated with multimodal video and audio models in a telehealth-style visit, request clinical data, and receive an AI evaluation that scores the encounter across 25 areas. | 10 Nov 2026 | Arizona State University | Diversified Consumer Services | United States |
Arizona State University's Research Computing team turned a donation of 229 Intel Gaudi2 servers (about 1,600 accelerators) into the AIR Platform, a campus AI service. Most nodes run bare-metal Kubernetes behind HAProxy, LiteLLM and AIBrix, serving open models through the Gaudi2 fork of vLLM to OpenWebUI, IDEs, Jupyter and CreateAI; ten more run batch jobs on the Sol supercomputer under Slurm. In the first half of 2026, 258 API-key holders sent more than 47 billion tokens. | 10 Nov 2026 | Arizona State University | Diversified Consumer Services | United States |
Anthropic used Claude Code to port Bun from Zig to Rust in less than two weeks. Rulebooks, adversarial agent reviews, compiler-driven fixes and an existing TypeScript test suite supported the migration; the shipped binary was 19% smaller on Linux and Windows and 2–5% faster in reported workloads. | 7 Nov 2026 | Anthropic | Software | United States |
Bloomberg's Automated Uplift initiative ran hybrid LLM-deterministic pipelines across thousands of repositories, generating 2,333 pull requests and 729 successful Python packaging and C++-to-CMake migrations, which the team estimates represent well over a thousand person-days of manual migration work. | 7 Nov 2026 | Bloomberg L.P. | Capital Markets | United Kingdom |
Slack ran 200+ automated agent-driven end-to-end test executions across three execution models (Playwright MCP, Playwright CLI, AI-generated Playwright tests) and two input formats (natural language, YAML) on two fixed workflows. MCP kept failure rates at 0-12% against 12-20% for the CLI; generated tests were fastest at about 3 minutes per run but failed about 48% of the time on the complex flow. Agent-driven runs cost $15-30 each, and the search-discovery flow used roughly 3.5M-7M tokens depending on harness and model. | 7 Nov 2026 | Slack | Software | United States |
Tencent's QQ Speed engineering team runs an AI-assisted R&D workflow built around reusable feedback loops: an LLM-based custom linter that self-repairs after each agent edit, CI pipelines that hand a failing lint stage back to the agent, narrow rule-based auto-approval, scheduled issue and conflict cleanup, and a preparation/execution/review split feeding findings back into team skills and memory. Reported scale is roughly 30 billion tokens a month and 10 to 20 concurrent agent workspaces per engineer, with failure modes including 200k-token context overflow and runaway sub-agent forks. The source is an edited conference-talk transcript, so figures are speaker-reported; a Wuhan municipal report of August 2026 places Tencent game teams working on QQ Speed in Wuhan, which is the basis for the city attribution. | 6 Nov 2026 | Tencent | Interactive Media & Services | China |
Tencent Cloud describes building CloudQ, an AI-native SRE agent that combines an operational knowledge graph with multi-agent orchestration for architecture visualization, impact analysis, root-cause investigation, change-risk control, inspection, capacity planning, and FinOps. The graph joins infrastructure topology, observability signals, change events, and operational knowledge. A Harness layer adds bounded agent execution, sandbox isolation, multi-cloud adapters, and online/offline evaluation with regression gates. The InfoQ article is an edited conference-talk report; its incident walkthroughs are illustrative and it provides no independently measured customer outcomes. | 5 Nov 2026 | Tencent Cloud | IT Services | China |
Revolut, the London-based digital bank, self-hosts an open-weight GLM coding model on 56 NVIDIA B200 GPUs to give about 2,000 engineers flat-cost, unmetered AI coding. Its AI operations lead reports a 60% cut in coding time and 70% in lead time after the switch in April 2026, a blended cost of about $0.30 per million tokens, roughly one tenth of its previous frontier-lab spend, and an 88% latency improvement from inference tuning on the same hardware. A sandbox with operation allow-lists keeps a human in the loop for EU banking supervision. | 4 Nov 2026 | Revolut | Financial Services | United Kingdom |
InfoQ reports Xiaohongshu scaled its OpenClaw-based Seal from a three-day demo to all-staff use in about a month. It describes cloud Kubernetes and Nex microVM sandboxes. The team reported Self-GC lowered daily token use 10–15% (about 20% at peak); Auto routing lowered model costs about 69%. | 4 Nov 2026 | Xiaohongshu (RedNote) | Interactive Media & Services | China |
The AI-IR lab at the University of Texas Medical Branch reimplemented 21 pretrained medical imaging models, including nnU-Net, TotalSegmentator, MedSAM2 and SAM 3, natively in MLX for Apple Silicon. Benchmarked against the original implementations on an Intel CPU and an NVIDIA RTX 4090 across public cases from 11 imaging datasets, the models matched outputs in 211 of 212 comparisons, and 19 of the 21 ran in under 16 GB of memory on a 64 GB Apple M1 Max, with a memory-saving setting cutting memory use by a median 49% at a median 13% processing-time cost. | 3 Nov 2026 | AI-IR Lab at University of Texas Medical Branch | Research Institution | United States |
PetroChina's Research Institute of Petroleum Exploration & Development built a shale oil large language model on Qwen2.5-72B and Qwen3-32B using a 21-billion-token domain corpus, QLoRA fine-tuning with dynamic context parallelism and a reflective RAG loop, reaching 86.35 on objective and 75.54 on subjective shale oil tests versus 77.32 and 12.12 for GPT-4o. A follow-on agent built on the Model Context Protocol orchestrates 23 geoscience tools for seismic processing and interpretation and scored 94.377 on a four-part tool-calling benchmark, above ChatGPT-5.4 and Kimi k2.6. | 3 Nov 2026 | PetroChina Research Institute of Petroleum Exploration & Development (RIPED) | Oil, Gas & Consumable Fuels | China |
Slack rebuilt how it serves the large language models behind Slack AI in four phases: models hosted in an escrow VPC on AWS SageMaker, a zero-incident migration to Amazon Bedrock, a hybrid of provisioned and on-demand capacity with spillover, and from early 2026 a multi-cloud setup adding Google Cloud Vertex AI. A routing layer picks a model per feature, with backup models and circuit breakers, lifting complex-reasoning quality about 10% and cutting latency about 67% on low-token workloads. | 2 Nov 2026 | Slack | Software | United States |
Showing 20 / 722 filtered rows