From Dev.to Community
Tech Articles
Curated developer articles, tutorials, and guides — auto-updated hourly
Tanay Joshi1d ago • 6 min read
Never lose a training run again: a checkpoint-and-resume playbook for ephemeral GPUs
▶ Prefer to play with it? There's an interactive version of this article where you can break things....
#machinelearning#python#mlops#learning
5 1
Sahajmeet Kaur4h ago • 7 min read
"AI Gateway vs API Gateway: They Solve Different Problems
We already had Kong running. Adding AI workloads on top of it made sense - until it didn't. Here's t...
#ai#devops#llm#mlops
2 0
Marcus Chen2d ago • 4 min read
Harvesting a regression test set from gateway logs with a plugin
TL;DR: Our eval sets went stale because a human wrote the test cases by hand once and never updated....
#mlops#llm#machinelearning#infrastructure
0 0
Elise Moreau2d ago • 4 min read
The SDXL VAE overflow that decoded black images in fp16
TL;DR: The SDXL VAE decoder pushes activations past 65504, the max value fp16 can hold, so the last....
#pytorch#computervision#machinelearning#mlops
1 0
Gabriel Henrique4d ago • 5 min read
Data Contracts in Production: Stop Trusting Your Upstream Sources
Your upstream data source changed a column type last night. Your pipeline ran at 2am, ingested...
#dataengineering#python#data#mlops
0 0
Marcus Chen1d ago • 4 min read
Benchmarking 5 LLM providers on one eval set, no SDK per vendor
TL;DR: We run a 1,200-case eval suite for enterprise agent automation at Nexus Labs. Comparing model...
#machinelearning#llm#mlops#devops
0 0
Marcus Chen2d ago • 4 min read
temperature=0 didn't make our LLM evals reproducible
TL;DR: We set temperature=0 and seed=42 and still got different eval scores on the same 800-prompt.....
#machinelearning#llm#mlops#infrastructure
0 0
Elise Moreau1d ago • 4 min read
Using the channels-last memory format reduced the latency of our conversation backbone by 22%
TL;DR: Switching our convolutional segmentation backbone to PyTorch's channels-last memory format cu...
#pytorch#computervision#machinelearning#mlops
1 0
Marcus Chen6d ago • 4 min read
Perplexity held flat after INT4. Task accuracy dropped 7 points.
TL;DR: We quantized a fine-tuned 14B agent model to INT4 with GPTQ. Perplexity moved 0.04. We almost...
#machinelearning#llm#mlops#pytorch
0 0
Elise Moreau6d ago • 4 min read
The seam our tiled upscaler left on every 4K product render
TL;DR: We tile high-res images through our upscaler because a full 4096×4096 pass blows past 24GB of...
#mlops#computervision#pytorch#machinelearning
0 0
SHIVAM UPADHYAY2d ago • 4 min read
Building a Self-Hosted MLOps Platform from Scratch with FastAPI, PostgreSQL, GCS, and Docker
Introduction Over the past few months, I set out to answer a simple question: What does...
#ai#devops#python#mlops
0 0
Suman Nath3d ago • 3 min read
If a 270M Model Already Worked, Why Did I Fine-Tune a 7B One?
Part 4 (finale) of a 4-part series. Three model sizes tied on the same task — so when does bigger ac...
#machinelearning#llm#mlops#ai
0 0
claire nguyen2d ago • 4 min read
Semantic caching our flaky-test summariser: 58% fewer LLM calls
TL;DR: Our internal flaky-test summariser at Buildkite was firing ~40k LLM calls a day, and most wer...
#sre#devops#llm#mlops
0 0
Mridul Nagpal1d ago • 3 min read
Machine learning in production: the model is the easy part
A model that scores 95% on your test set feels like the finish line. Then you ship it, and you find....
#ai#machinelearning#mlops
3 1
Fernando Azevedo1d ago • 11 min read
ML Observability on EKS: Logs, Metrics and Tracing Head-to-Head
Technical analysis comparing the leading observability strategies for ML workloads on EKS: Fluent Bi...
#dataplatforms#eks#mlops#observability
0 0
Sanskriti Harmukh1d ago • 3 min read
Deploying MLflow Open-Source Machine Learning Experiment Tracking on Ubuntu 24.04
MLflow is an open-source platform for managing the machine learning lifecycle — experiment tracking,...
#mlops#docker#devops#ai
6 0
Renato Marinho11h ago • 4 min read
Stop building custom wrappers for your ML models.
I spent three days last month building a specialized API wrapper for a simple Scikit-learn model. No...
#ai#mlops#python#productivity
0 0
Sahajmeet Kaur5d ago • 7 min read
Portkey Alternative: I Switched Away from Portkey. Here's the Honest Reason Why.
I ran Portkey in production for six months and genuinely liked it — until the acquisition, the cost ...
#ai#llm#devops#mlops
1 0
Sahajmeet Kaur1d ago • 8 min read
What Is an Agent Gateway? (And Why Our AI Gateway Stopped Being Enough)
We had an AI gateway running fine for six months. Then we added agents. Here's what broke - and why ...
#ai#agents#devops#mlops
2 0
Marcus Chen4h ago • 4 min read
Position bias in LLM-as-judge flipped 18% of our verdicts
TL;DR: Position bias in LLM-as-judge means the model favors whichever answer it reads first. We...
#machinelearning#llm#mlops#aiengineering
0 0
Naveen Malothu6d ago • 2 min read
Exploring Midjourney Medical: Revolutionizing Healthcare with AI
Discover the potential of Midjourney Medical in revolutionizing healthcare with AI-generated medical...
#ai#machinelearning#mlops
0 0
James Sanderson3d ago • 3 min read
Scaling Industrial Intelligence: Architectural Patterns from a Machine Learning Development Company
While traditional software engineering relies on static, rule-based logical structures, modern...
#ai#mlops#machinelearning#deeplearning
0 0
Scott McMahan2d ago • 2 min read
Machine Learning Engineer vs Data Scientist: Understanding the Difference
Artificial intelligence teams often include both Data Scientists and Machine Learning Engineers,...
#ai#machinelearning#datascience#mlops
0 0
joseph quesada2d ago • 8 min read
Real-time AI: Build Your Live Data Stack for Performance
Unlock real-time intelligence for your AI campaigns and applications. This post details how to build
#aiinfrastructure#realtimedata#dataengineering#mlops
0 0