All technologies A2C Agents Alembic Amazon SageMaker Anomaly detection Apache Beam AWS AWS Batch AWS Glue AWS Lambda Causal inference Chroma CI/CD Clark-Scarf Claude Clustering Contrastive learning Cross-attention CUDA C++ Differential privacy Distributed training Docker Domain adaptation DPO DQN EC2 EconML Edge inference Elasticsearch Fashion-CLIP FastAPI Federated learning Fine-tuning Flink Flower GCP Generalized-EBs GitHub Actions GPU optimisation Grafana Grounding DINO Horovod Human-in-the-loop Influencer marketing InfoNCE Istio Jenkins Kafka KEDA Kubeflow Kubernetes LangChain LoRA Marketing Marketing budget allocation MCP MILP MLflow MLOps Multimodal retrieval Neo4j Observability ONNX Runtime Opacus Operations research Optuna OutfitTransformer PM4py PPO Process mining Prometheus pytest Python PyTorch QDA Quantization RAG Ray Ray Tune Recommender systems Redis Reinforcement learning Responsible AI RLHF Scala SciPy / HiGHS SQLAlchemy SQLite Streamlit Submodular optimization Supply chain TensorFlow.js TensorRT-LLM Terraform Triton Inference Server VAE Variational autoencoder Vector search Vision-language models vLLM WASM Weakly supervised learning XGBoost Zero-shot classification
Fintech / Credit risk 2026
A causally-audited reinforcement-learning pipeline for loan/insurance offer optimization, built on real BPI Challenge 2017 event data.
Offers in demand model 42,821
Events / applications 193,849 / 31,509 PM4py EconML Apache Beam Kubernetes Kubeflow Process mining Causal inference View case study E-commerce / Agentic AI 2026
An agentic shopping assistant with four retrieval strategies as tools, and no capability to create a billable order without human sign-off.
Agent tools exposed over MCP 8
Retrieval strategies 4 LangChain MCP Elasticsearch Neo4j Vector search SQLAlchemy Alembic View case study AI platform / Infrastructure 2026
Qwen2.5-3B-Instruct benchmarked end to end , distributed DPO training, hand-written CUDA kernels, a three-way inference-engine comparison, Triton serving and Kubernetes autoscaling.
Fused CUDA kernel speedup 2.43x
Throughput gap vs. ONNX Runtime GenAI 7.2x DPO RLHF LoRA Ray Horovod CUDA C++ ONNX Runtime View case study Supply chain / Operations research 2025
RL agents that match classical optimal stock policies on fixed lead-times and outperform them when lead-times vary.
Optimality gap vs C-S (fixed LT) 0.02%
Lift vs alternatives (varying LT) +4.1% Reinforcement learning DQN A2C PPO Supply chain Operations research Clark-Scarf View case study Automotive / Supply chain quality 2025
A hybrid deep-learning and classic-ML anomaly detection platform that flags supply-chain irregularities and recommends classification.
Projected annual savings €1.6M+
Model families combined 4 Anomaly detection Variational autoencoder VAE Clustering XGBoost QDA Weakly supervised learning View case study 