Blog
Builder Center2026Language models solved only 34% of SWE-InfraBench's infrastructure-as-code tasks in one attempt. We ran Kiro over all 100 and found that a full agent loop with Sonnet 5 reaches 82%, with partially correct answers converging once the agent can re-run the tests itself.
AWS Blog2026Defects that clear compilers, static analyzers and code review still surface during vehicle integration testing, the most expensive place to find them. We built a multi-agent system that detected 30+ high-priority findings on production code.
Builder Center2026We benchmark Task Arithmetic, TIES, DARE, and Model Soups for merging Qwen3.5 experts across five domains. The best expert training duration turns out to depend on the merging method, with sparsification-based methods peaking well past the validation optimum.
AWS Blog2026We orchestrate a team of specialized agents that runs medical, legal, and regulatory review of pharmaceutical content. Promotional material is checked against scientific references, regulatory codes, and approved claims, cutting review cycles by up to 90%.
Builder Center2026An agentic pipeline on Amazon Bedrock AgentCore that checks medical marketing content against compliance guidelines. We look at agent design, tool integration, and what it takes to run the workflow at scale.
Builder Center2026AgentCore Deep Research is an open-source application for multi-source research workflows. We cover deploying it, customizing the research agents, and adapting the pipeline to new data sources.
AWS Blog2026Maritime analysts drown in isolated alerts that carry no context on their own. Together with Windward, we built an agentic system that correlates them into enriched anomaly reports, shifting effort from data collection to decisions.
AWS Blog2026Undocumented mainframe code is the knowledge-loss problem that stalls enterprise migration. We built agentic pipelines on Amazon Bedrock that generate documentation from legacy sources at Toyota Motor Europe.
Builder Center2026Self-reflection lets a model critique and revise its own output before returning it. We apply it at inference time on Amazon Bedrock and measure the accuracy gain against the added cost and latency across reflection depths.
AWS Blog2026We built MACROS, a system on Amazon Bedrock with Lambda and Step Functions that reviews medical articles against guidelines and proposes revisions. The proof of concept reached 80% accuracy and over 90% recall, cutting per-guideline review from hours to minutes.
AWS Blog2026We paired a fine-tuned Qwen3-32B with Claude Sonnet orchestration over Amazon Bedrock and MCP to build a multi-model coding assistant. Fine-tuning on 7,000 annotated functions cut C++ architecture compliance violations to 0.02 against generic models.
AWS Blog2025AUMOVIO's vSDV Composer pairs virtual ECUs with an Amazon Bedrock engineering assistant trained on 20+ years of automotive source code. We surface compliance violations earlier in development, with a reported 50%+ acceleration in delivery.
AWS Blog2025We ship an end-to-end document processing application that extracts structured data from PDFs and emails. Three extraction paths are compared: managed BDA, custom Bedrock models, and Textract OCR paired with an LLM.
AWS Blog2025RAG systems still answer confidently when the retrieved context does not support the answer. We add a detection layer that flags those cases, comparing methods such as LLM-as-judge and semantic similarity on accuracy against cost.
AWS Blog2024We wire up a system on Amazon Bedrock that drafts medical marketing content from scientific references and revises it through expert feedback loops. Textract handles document parsing and Amazon Translate the localization, cutting drafting from weeks to hours.
AWS Blog2023When retrieval returns a weak answer, most RAG systems reply anyway. We add a clarification tool so the system asks a follow-up question instead, combining Amazon Kendra retrieval, LangChain orchestration, and Amazon Bedrock.
Self-published2023Principal Component Analysis, built from first principles in Python and NumPy and wrapped in a scikit-learn-style class. We work through the covariance matrix, eigendecomposition, and component selection on a toy dataset.
Self-published2023We implement K Nearest Neighbors from first principles in Python and NumPy as a reusable class object. Distance computation, neighbor selection, and majority voting are each built up and tested on a toy dataset.
Self-published2022Fine-tuning a pre-trained network works better when different layers move at different speeds. We set up discriminative learning rates with PyTorch parameter groups, so early layers retain what they learned while later layers adapt faster.
Self-published2021This post adapts test-time augmentation, a computer vision technique, to tabular data. We benchmark strategies for numeric and categorical features and measure how much accuracy they actually buy at inference.
Self-published2021Sometimes the useful signal sits in a hidden layer rather than the output. We compare three ways to read those activations in PyTorch, from forward hooks to slicing the module list, and use them to pull out embeddings.
Self-published2021Spreadsheets stop scaling once an experiment log runs to hundreds of rows. We move tracking onto Neptune.ai across a Kaggle competition workflow, logging metrics, parameters, and artifacts.
Self-published2021We compute channel-wise mean and standard deviation across an image dataset for normalization. The implementation batches through the data in PyTorch, so it never holds the full dataset in memory.
Self-published2021Large models often will not fit a useful batch size into GPU memory. We use gradient accumulation to reach a large effective batch, walking through the modified training loop and its impact on learning rate scaling and batch norm.
Self-published2020This post covers training PyTorch models on Cloud TPUs with PyTorch/XLA, on both Kaggle and Colab. We go through the XLA device model, multi-core distributed training, and the changes needed to port an existing GPU pipeline.
Self-published2020When labeled training data is scarce but unlabeled test data is plentiful, the test set is still useful. We pre-train on surrogate labels derived from it, then fine-tune on the real labels to lift network performance.
Self-published2020We forecast retail demand with gradient boosted trees in our Data Mining Cup 2020 solution. Optimizing a profit-based objective rather than plain forecast error aligns the model with the underlying inventory cost and revenue trade-offs.
Self-published2020A convolutional network that grades diabetic retinopathy severity from retina photographs, built for a Kaggle competition and the Udacity ML Engineer program. We cover retina-specific preprocessing, augmentation, and the ordinal grading setup.
Self-published2020This post introduces the fairness R package for computing and visualizing algorithmic fairness metrics. We walk through demographic parity, proportional parity, and equalized odds, and show how the three diverge on the very same classifier.