> Markdown version of [/videos/154-machine-learning-for-software-developers-and-knitters?t=790](https://www.wearedevelopers.com/videos/154-machine-learning-for-software-developers-and-knitters?t=790). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Machine Learning for Software Developers (and Knitters)

Can an AI reverse-engineer a knitting pattern? Step into the Software 2.0 paradigm and learn how to build AWS computer vision pipelines without a data science degree.

- **Speakers:** Kris Howard
- **Event:** WeAreDevelopers LIVE
- **Published:** May 26, 2021
- **Duration:** 49:45
- **URL:** https://www.wearedevelopers.com/videos/154-machine-learning-for-software-developers-and-knitters

## Summary

The transition toward the "Software 2.0" paradigm underscores a fundamental industry shift: for the last 40 years we have programmed computers, but going forward, we will train them. This requires traditional software developers to adapt to new patterns, moving away from explicit instructions toward systematically handling probabilistic confidence scores rather than absolute binary outcomes. To demonstrate that ML is accessible without deep calculus or data science expertise, this session explores a real-world computer vision experiment: training an image classification model to reverse-engineer knitting patterns from photographs.

Building an automated, event-driven data pipeline using Amazon SES, AWS Lambda, Amazon S3, and Amazon DynamoDB allowed for scalable crowdsourcing of knitted swatch images. However, crowdsourced datasets demand significant hygiene. Developers quickly learn that 80% of ML work is data cleaning; ambiguous 2D photos and regional naming variations forced this project to manually scope down from a multi-class identification problem to a simpler binary classification approach. Using Amazon SageMaker Ground Truth for labeling highlighted the absolute necessity of clean inputs, while augmenting the dataset—via rotation, grayscale conversion, and introducing "clutter" (random objects like cats and dinosaurs)—ultimately prevented the model from overfitting to domain-specific textures.

In a comparative evaluation of the AWS ML stack, the project tested custom model training on Amazon SageMaker against Amazon Rekognition Custom Labels. While SageMaker, paired with hyperparameter tuning, yielded high accuracy, it required managing endpoint configurations and intermediary serverless inference architectures. Conversely, Rekognition Custom Labels provided an out-of-the-box UI that achieved ~99% accuracy with drastically reduced configuration effort. This illustrates a vital structural trade-off between control and cost: top-tier AI services democratize machine learning for application developers but limit granular architectural control, making batch inference strategies crucial for cost optimization. Ultimately, the integration of these tiered services allows teams to seamlessly infuse AI logic into existing workflows.

**Keywords:** software 2.0 paradigm, probabilistic confidence scores, amazon sagemaker integration, amazon rekognition custom labels, sagemaker ground truth labeling, serverless data pipelines, crowdsourced dataset cleaning, image classification models, training data cluttering, hyperparameter optimization, AWS ML stack architectures, serverless inference APIs, batch inference cost optimization, binary classification scoping

## Chapters

1. **Approachable machine learning concepts for software developers** (00:17) — Bypassing complex mathematical theory allows developers of any background to conceptualize and integrate basic machine learning capabilities.
1. **Commercial applications of modern artificial intelligence and machine learning** (03:08) — Analyzing advanced implementations in biology and delivery operations illustrates how predictive algorithms solve uniquely complex logistics challenges.
1. **Transitioning from deterministic programming to probabilistic machine learning frameworks** (09:43) — Adapting to probability-based outputs rather than discrete logic transforms how traditional engineers must structure conditional branching.
1. **Exploring the tiered architecture of modern machine learning stacks** (13:10) — Breaking down the cloud artificial intelligence stack enables practitioners to match deployment options directly to their technical experience levels.
1. **Translating physical object reverse engineering into image classification problems** (15:28) — Approaching physical structure identification as an image classification task creates a pathway for digitizing real-world textures.
1. **Automating data collection and managing crowdsourced training image sets** (21:03) — Establishing automated cloud storage pipelines mitigates the overwhelming file cleanup tasks generated by crowdsourcing raw multi-media data.
1. **Curating accurate training labels via managed workflows and public workforces** (26:12) — Designing restrictive categorization workflows for human annotators minimizes the systemic inaccuracies caused by subjective visual interpretations.
1. **Executing custom image classification training jobs in managed environments** (29:49) — Executing structured validation cycles within comprehensive cloud environments simplifies the notoriously difficult procedure of deploying scalable inference endpoints.
1. **Improving model inference accuracy via data augmentation and parameter tuning** (33:29) — Introducing irrelevant visual clutter and optimizing hyperparameters resolves the severe false positive rates commonly found in early vision models.
1. **Evaluating tradeoffs between fully managed artificial intelligence and custom pipelines** (39:03) — Evaluating pre-packaged vision classification services against flexible scientific platforms highlights the crucial balance between administrative control and developmental speed.

## Related Moments

- [Introduction to prototyping and building practical AI applications](https://www.wearedevelopers.com/videos/1010-bringing-the-power-of-ai-to-your-application) (from "Bringing the power of AI to your application.")
- [Uncovering the hidden technical debt in machine learning](https://www.wearedevelopers.com/videos/392-mlops-what-s-the-deal-behind-it) (from "MLOps - What’s the deal behind it?")
- [Introduction to artificial intelligence driven development](https://www.wearedevelopers.com/videos/347-mlops-and-ai-driven-development) (from "MLOps and AI Driven Development")
- [Audience questions on practical machine learning operational strategies](https://www.wearedevelopers.com/videos/262-is-my-ai-alive-but-brain-dead-how-monitoring-can-tell-you-if-your-machine-learning-stack-is-still-performing) (from "Is my AI alive but brain-dead? How monitoring can tell you if your machine learning stack is still performing")
- [Designing an automated machine learning deployment blueprint](https://www.wearedevelopers.com/videos/1050-the-road-to-mlops-how-verivox-transitioned-to-aws) (from "The Road to MLOps: How Verivox Transitioned to AWS")
- [Implementing continuous delivery architecture for machine learning](https://www.wearedevelopers.com/videos/178-industrializing-your-data-science-capabilities) (from "Industrializing your Data Science capabilities")

## Related Articles

- [How machine learning can help us tell fact from fiction](https://www.wearedevelopers.com/magazine/509-how-machine-learning-can-help-us-tell-fact-from-fiction)
- [MLops – Deploying, Maintaining And Evolving Machine Learning Models in Production](https://www.wearedevelopers.com/magazine/115-mlops-deploying-maintaining-and-evolving-machine-learning-models-in-production)
- [MLOps – What’s the deal behind it?](https://www.wearedevelopers.com/magazine/125-mlops-what-s-the-deal-behind-it)
- [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development)

## Related Jobs

- [Staff, Machine Learning Engineer (L4)](https://www.wearedevelopers.com/jobs/ext/1202639-staff-machine-learning-engineer-l4) at **Twilio**
- [Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/588393-machine-learning-engineer) at **Twilio**
- [Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/1355348-machine-learning-engineer) at **TWILIO**
- [Principal Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/1706410-principal-machine-learning-engineer) at **Almedia**
- [Data Scientist](https://www.wearedevelopers.com/jobs/ext/1351648-data-scientist) at **Almedia**
- [Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/1597388-machine-learning-engineer) at **ZEISS Group**