Shenghu Yang

Insightful guide on building multi-agentic meta-tooling systems using AWS Strands SDK . It's fascinating to see how Strands simplifies the orchestration of specialized AI agents, leveraging models like Claude 4 Sonnet and Opus to tackle complex financial analyses. The ability to seamlessly integrate multiple agents for tasks like fundamental, technical, and market analysis showcases the potential of collaborative AI systems. How might this approach transform other industries beyond finance? https://lnkd.in/evcKaYx8 #MultiAgentSystems #AWSStrands #GenerativeAI

6

"Lithic, a card processing platform, now integrates with Visa's ALM via the VCPE API for enhanced card program delivery and economics." See this and many other announcements in this week's issue of the API Changelog newsletter: https://lnkd.in/dAPJm-fs

4

Quick AI Cost Comparison — July 2025 If you’re building AI-powered products, here’s a rough pricing snapshot (per 1K tokens or per image): * Google Gemini is extremely cost-effective. * Cloudflare offers low prices + model flexibility. * Anthropic and OpenAI offer premium quality at a higher price point — useful when accuracy matters most. You can stretch your budget by fine-tuning smaller models or using open-source with smart prompting. ⚠️ Note: These numbers are approximations. Actual costs vary based on: - Input vs. output tokens (e.g. OpenAI charges more for output) - Image size and resolution (some models tokenize image content) - Tiered or discounted pricing at scale - Whether you’re using hosted endpoints (like Cloudflare Workers AI) or cloud platforms (like Vertex AI)

11

1 Comment

The industry doesn't need more people who can "write code." LLMs can do that now. The industry needs people who can: 1. Identify bottlenecks before they happen. 2. Structure systems for 10x growth. 3. Architect AI-native workflows. The Republic Sale ends in 48 hours. 70% OFF HLD + LLD + AI Combos : https://lnkd.in/g5rUq944

3

1 Comment

YellowDog.ai just set a 10x benchmark uplift in scale computing, delivering 40,000 tasks per second (TPS) and managing 100,000 compute nodes in the cloud. What's even more interesting, that's 2x IBM Symphony and opens an intriguing pathway for these until-know closed/captive systems.

11

An agent has no consistent definition. You can think of an agent as an autonomous or semi-autonomous system that can take actions on behalf of the user in a given environment, state, and make decisions or take actions that can accomplish certain tasks along a given time-frame, either by calling tools, dynamically selecting the next action to take or use a deterministic "workflow" (systems where LLMs and tools are orchestrated through predefined code paths). Using an Agent framework that gives you the ability to built systems in both, a reliable and a dynamic way can accelerate your agent development journey using agent abstractions. This means building Agentic systems that can reliably call tools, store memory (episodic/semantic/procedural), have comprehensive logging and observability, human in the loop workflows and the ability to build various multi-agent patterns flexibly based on your use case. A successful Agentic system in production is usually a combination of both, dynamic and predictable/reliable multi-agent systems. Strands Agents SDK gives you exactly these capabilities by treating each “agent” as a combination of a foundation model plus a suite of tools. You define a prompt and register your tools (decorated functions) in code, then Strands handles reasoning→planning→tool-execution cycles, local testing, and cloud deployment (ECS, Fargate, Lambda, EC2), along with support for all other agent abstractions provided above. Excited to share a hedge-fund analyst multi agent system: This uses the newest Anthropic's Claude 4 Sonnet/Opus that powers the Lead Analyst Agent, routing incoming queries to specialized sub-agents for fundamental, technical, and market analyses. Each specialist is wrapped as a callable tool (using the “agents-as-tools” multi-agent pattern), so the orchestrator never has to implement domain logic itself and can handoff the task to an agent as a tool. For sensitive operations (insider lookups), we utilize a HITL approval step that halts execution until a human grants consent. We also use meta-tooling that enables the Lead Analyst to generate, load, and invoke new custom tools at runtime—whether it’s a portfolio beta calculator or a pricer—without redeployment. Strands also embeds observability (Langfuse) and OpenTelemetry tracing so you can trace reasoning events, tool invocations, errors, and end-to-end workflows in real time. View more information on the code implementation here: https://lnkd.in/gJmwVyGi Code implementation: https://lnkd.in/gzTtJvJq Thanks to 🏄‍♂️ Cagatay Cali for being a reviewer/collaborator on this! Feel free to try it out and reach out with any questions/ideas. #aws #agenticAI #strands #agents #generativeAI

60

6 Comments

One AI-enabled company I spoke with organizes its engineering team like this: → 9 two-person engineering pods → 1 product manager → 1 designer Compare this to a legacy feature team: → ~ 6 engineers → 1 product manager → 1 designer On well-defined greenfield projects, these two-person AI pods ship a similar amount of software as a six-engineer legacy team. To run 9 legacy feature teams: 72 people. To run 9 AI pods: 20 people. That’s ~3.6X more software delivery per headcount. To make this work you need: → Engineers with the judgment to operate independently → AI engineering best practices → AI-enabled product managers and designers (or else they become the bottleneck) When a company tells me they’ve adopted AI but haven’t seen meaningful gains, it’s usually because their team structure hasn’t changed. AI changes the optimal team size. The biggest gains come from smaller, more autonomous teams with less coordination overhead. I’m curious – has AI changed the structure of your team? If so, how?

31

18 Comments

How Attentive uses Gatling and Datadog to achieve 360° observability at scale For teams operating at scale, performance issues are rarely clear-cut. The real challenge is understanding whether traffic is flowing, where it slows down, and why. In an upcoming live session, Bian Jiang, Principal Software Engineer on Attentive's Developer Experience team, will walk through how Attentive connects Gatling and Datadog to gain end-to-end visibility during high-scale load testing. We’ll explore: - How Attentive correlates generated load with real system behavior - Why observability is critical to trusting load test results - Lessons learned testing systems under real constraints - How Datadog and Gatling together close blind spots at scale 📅 January 22, 2026 - 6PM CET / 12PM EST 🎙 Bian Jiang, Principal Software Engineer, Attentive 🗃️ Diego Salinas, Enterprise Content Manager, Gatling 🗃️ Heather Thacker, Developer Relations, Gatling 🔗 Register here: https://lnkd.in/eqvEuMhB

18

2 Comments

𝗛𝗼𝘄 𝗠𝗖𝗣 𝗦𝗲𝗿𝘃𝗲𝗿𝘀 𝗠𝗮𝗻𝗮𝗴𝗲 𝗦𝗲𝘀𝘀𝗶𝗼𝗻𝘀 𝘄𝗶𝘁𝗵 𝗟𝗟𝗠𝘀 𝗮𝗻𝗱 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗪𝗶𝗻𝗱𝗼𝘄𝘀 Managing sessions in LLMs isn’t just about keeping track of prompts. It is about orchestrating memory, context, and tools seamlessly across multiple agents. That’s where MCP servers come in. Here’s the flow I captured visually → Context Window as Working Memory: Holds system prompt, user inputs, assistant responses, tool calls, and rolling context → MCP Servers: Each server maintains its own session state, ensuring consistency across interactions → System Prompt + Tools + Code Loops: Work together to power agentic behavior while grounding outputs in real tasks → LLM Integration: The context is constantly summarized and updated before being sent to the model This structure ensures scalable, multi-session orchestration, which is critical for enterprise AI where multiple tasks and tools interact simultaneously.

149

14 Comments

🚀 Still sending raw JSON to your LLMs? You’re probably wasting tokens — and money — every single request. JSON is great for software. Deterministic. Verbose. Explicit. LLMs don’t need that much ceremony. Every {, ", and nested key → becomes extra tokens → becomes higher latency → becomes higher API bills 💸 That’s why I started using TOON (Token-Oriented Object Notation). Same structure. Same meaning. Way fewer tokens. Example from a customer review dataset: • JSON: 412 characters • TOON: 154 characters That’s ~60% less formatting noise. What you get instead: ⚡ Faster responses 💰 Lower inference costs 🧠 More room for actual context Important caveat: ❌ JSON isn’t “bad” ✅ It’s just not optimized for LLMs When you’re building agents, RAG pipelines, or structured prompts, input format becomes a performance decision — not a style choice. AI engineering isn’t only about bigger models. Sometimes it’s about sending smarter input. 👇 Curious: are you optimizing prompt formats yet, or still only prompts? #LLM #GenerativeAI #AIEngineering #PromptEngineering #AIAgents #SoftwareEngineering #DataEngineering

29

7 Comments

The power of focused tiger teams In 2000, eBay cut the pricing for its in-house payment system BillPoint, posing an existential risk to PayPal. In response, David O. Sacks pulled together a tiger team, constituted of high performers across functions. Over a period of a few months, this team executed on numerous counteroffensive maneuvers that stopped BillPoint from growing. I’ve seen similar cross functional tiger teams successfully tackle exceptionally hard - existential - problems time and again, across companies. It works. Startup CEOs: when you face a crisis, don’t go it alone and carry the burden on your shoulders. Leverage the best people on your team. Create a tiger squad and task them with solving the problem. It really works. Ensure that the squad is comprised of high performers with a diverse set of functional skills and perspectives. This will enable the team to attack the problem from multiple angles simultaneously, identify blind spots you might have missed, and generate unconventional solutions that wouldn’t emerge from siloed thinking.

59

7 Comments