DATA Pill feed

DATA Pill #191 – NVIDIA Agents, GitHub Copilot SDK, Lambda ML Jobs & the Next Two Years of Engineering

ARTICLES

Multi-Agent Warehouse AI: A Command Layer for Operational Excellence and Supply-Chain Intelligence | NVIDIA Developer Blog | Tarik Hammadou and Jeremy Coupe | 8 min | Agentic Systems & Industrial AI
NVIDIA presents a real-world multi-agent architecture for warehouse operations, where specialized agents coordinate planning, perception, optimization, and execution through a shared command layer. The post shows how agent hierarchies, event-driven orchestration, and simulation-backed decision loops enable resilient, scalable supply-chain intelligence.
How to Deploy ML Jobs on Lambda Cloud with SkyPilot | Lambda AI | Cody Brownstein | 3 min | ML Infrastructure
This walkthrough shows how to use the open‑source SkyPilot framework to run ML training and inference jobs on Lambda Cloud. SkyPilot abstracts GPU provisioning and scheduling: you define resources in a YAML file and SkyPilot launches the job on spot or dedicated GPUs across multiple regions. The post provides example scripts for launching a Hugging Face training run and explains how SkyPilot handles retrying on preemptible instances.
Build an Agent into Any App with the GitHub Copilot SDK | GitHub Blog | Mario Rodriguez | 6 min | Developer Platforms
GitHub announces a Copilot SDK that lets developers embed agentic capabilities directly into their applications. The SDK provides a flexible CLI and HTTP API for building domain‑specific agents that can run tasks, query codebases and integrate with third‑party services. Examples include using the CLI to write migration scripts, generating changelogs from commit history and exposing custom commands via gh agent
The Next Two Years of Software Engineering | Addy Osmani | 10 min | Future of Work
Google’s Addy Osmani maps two scenarios for developers through 2026. In one, AI coding agents automate most entry‑level work, causing junior hiring to contract; a Harvard study found that when companies adopt generative AI, junior developer employment can drop by ~9‑10%. In the other, AI unlocks demand as every industry embeds software and automation, creating new “AI‑native” entry‑level roles. Osmani urges juniors to become AI‑proficient—using tools like Cursor, Antigravity and Claude Code—while building skills AI can’t replace (communication, problem decomposition, domain knowledge). Senior engineers, meanwhile, must learn to prompt and validate AI output and be ready to mentor, because the best developers will be those who know when AI is wrong
How Thomson Reuters Built an Agentic Platform Engineering Hub with Amazon Bedrock AgentCore | AWS ML Blog | Naveen Pollamreddi, Seth Krause, Pratip Bagchi, and Sandeep Singh | 7 min | Platform Engineering
Thomson Reuters shares how it created a platform engineering hub where internal teams build and orchestrate AI agents using Amazon Bedrock AgentCore. The hub provides a catalogue of secure, governed agents that automate tasks like infrastructure provisioning, compliance checks and knowledge base search. By standardising prompts, tool access and observability, the hub accelerates adoption while meeting legal and security requirements.

NEWS

Voice‑moderation startup Modulate introduces the Ensemble Listening Model (ELM), an AI architecture designed to improve accuracy, transparency and cost for real‑time voice moderation. ELM combines multiple specialized models in an ensemble that can transcribe, classify and detect harms in voice chat with lower latency and configurable safety thresholds. Modulate says ELM reduces false positives while allowing customers to tune models for different contexts.
UK bank Lloyds announces an AI Academy aiming to train its entire workforce—around 60 000 employees—in practical AI skills. Courses will cover generative AI, automation and data literacy, with specialized modules for engineers, product managers and customer‑facing staff. Lloyds says the academy will help embed responsible AI across its business and prepare employees for a future where AI augments most roles

TUTORIALS & BOOKS

Prompt Engineering for AI Models (Full Course) | Simplilearn | 2 h course | Prompt Engineering
A comprehensive introduction to prompt engineering covering how to craft effective prompts, use context windows, chain prompts for complex tasks and debug model outputs. Real examples and live demos help students practice constructing prompts that elicit accurate, useful responses.
AI Bootcamp Playlist | Dave Ebbelaar | Multi‑part series | AI Fundamentals
A curated playlist of lectures on AI fundamentals, including machine‑learning basics, deep learning, deployment and ethical considerations. Ideal for learners seeking a structured introduction to AI in 2026.

CONFS, EVENTS, WEBINARS & MEETUPS

Build Data & AI Literacy 2026 – People, Skills & Tools | Nina Stefels, Rozaliya Khafizova | Xebia | February 10 | Webinar
An interactive workshop on developing data & AI literacy across organisations. Topics include identifying skill gaps, creating training programmes, selecting the right tools and fostering a culture of responsible AI adoption.
Data & AI Warsaw Summit 2026| Warsaw | April 2026
A two‑day conference covering data engineering, analytics and AI. Speakers from global tech firms share case studies on lakehouse architectures, streaming, feature stores and agentic applications. Use code datapill10 for a 10% discount on tickets.

PINNACLE PICKS

Your last edition top picks:
Alternatives to MinIO for Single‑Node Local S3 | Robin Moffatt |rmoff.net | 6 min | Data Engineering

With MinIO pivoting away from open source, Moffatt evaluates alternatives for a simple, S3‑compatible storage layer. He lists essential criteria—Docker support, S3 API compatibility, simplicity and active community—and surveys options like SeaweedFS, Cloudflare R2, Zenko and others, comparing trade‑offs

A Realist’s Guide to Hybrid Mesh Architecture (Part 1): Single Source of Truth vs Democratisation | Xebia | XiaoHan Li | 9 min | Data Architecture

The Author argues that neither a pure data mesh nor a fully centralised warehouse fits enterprise realities. She proposes a hybrid hub‑and‑spoke pattern where domain teams own their data products but connect through a shared integration layer and platform guardrails. This preserves a single source of truth while still allowing local autonomy

Data Pipelines with Apache Airflow, Second Edition| Julian de Ruiter, Ismael Cabral, Kris Geusebroek, Daniel van der Ende, Bas Harenslak | Data Orchestration

This updated Airflow manual teaches readers how to build reliable, scalable pipelines with the latest Airflow features. Authors Bas Harenslak and Julian de Ruiter draw on extensive engineering experience and contributions to Airflow’s codebase.
_____________________
Have any interesting content to share in the DATA Pill newsletter?
2026-01-23 08:29