Interview with Abhinav Shashank on Innovaccer's data process and products

🌐 Unlocking the Power of APIs: What Are They and Why Do They Matter? Imagine ordering your favorite coffee without knowing how it’s brewed—APIs work the same way! An Application Programming Interface (API) is the bridge that lets apps talk to each other, delivering data or services seamlessly. How are APIs used? Developers: Build cool features, like embedding Google Maps or pulling live data into apps. Businesses: Automate workflows, connect tools (think CRM + email platforms), or create innovative services. You: Ever logged into a site with your Google account or shared a post directly to social media? That’s an API at work! APIs are the unsung heroes of our connected digital world, driving efficiency and innovation. 🚀 💡 Pro tip: Want to supercharge your projects with AI? Check out xAI’s API at https://x.ai/api for cutting-edge possibilities! What’s your favorite API-powered tool? Share below! 👇 #TechSimplified #APIs #Innovation #Developers

⚡ Forge Insight 4: Data Activation Without the Dev Queue Real-time marketing, no engineering bottlenecks. With data warehouse syncs and zero-copy Canvas triggers, marketers can activate journeys directly from backend or cloud tables without replication or waiting for developers. This is a game-changer: real-time data drives faster reactions, more agile campaigns and tighter alignment between marketing and engineering. At SCHMACK | Growth Agency 🚀, we’re helping clients reorganise around this agility - turning live data into live campaigns. Real-time marketing isn’t a nice-to-have anymore. It’s the new standard. Braze #BrazeForge #Martech #DataDrivenMarketing #BrazeAI #CRMStrategy

Automate an end-to-end AI newsletter with a ready-to-run n8n template — from raw markdown and tweets to a polished, Axios-style email. This template wires together S3/R2 ingestion, LangChain-powered story selection, per-story content aggregation and source scraping, constrained LLM prompts for tightly formatted segments, and an approval loop via Slack. It produces a finished markdown newsletter and uploads it to Slack so your team can review and publish quickly. Highlights: - Ingest daily .md files and tweets from a data-ingestion bucket - Use structured LLM outputs to pick and justify the top 4 stories (with content identifiers) - Resolve story identifiers, download original content, and scrape linked sources and images - Generate Axios-style sections, intro, and "other top stories" with strict prompt schemas - Auto-create subject line + pre-header and support Slack review/edit & approval - Output a complete markdown file ready to publish or distribute Who should use it: newsletter teams, content ops, product teams, and any developer who wants a reproducible, auditable pipeline for publishing AI-focused email digests. Why it matters: it reduces manual curation overhead, enforces consistent formatting and source-tracking, and makes LLM-driven content production reliable and editable by teams. Link to the template is in the first comment — fork it in your n8n instance, plug in your S3/R2, LLM, internal API, and Slack creds, then iterate. If you want a walkthrough or help adapting it, drop a comment or DM me. Template link in the comments section. #n8n #n8nWorkflow #WorkflowAutomation #AIAutomation #EmailMarketing #Newsletter #NewsletterAutomation #ContentAutomation #MarketingAutomation #LangChain #LLM #GenerativeAI #GoogleGemini #Anthropic #AWS #AmazonS3 #CloudflareR2 #Slack #WebScraping #OpenSource

Today I've been testing Parlant, a new open-source library for Python to build a small assistant for my DataBites newsletter ⚙️ 𝗠𝘆 𝗴𝗼𝗮𝗹 🎯 Help readers explore topics, search past issues, and get concise answers (𝘸𝘪𝘵𝘩𝘰𝘶𝘵 𝘵𝘩𝘦 𝘢𝘨𝘦𝘯𝘵 𝘥𝘳𝘪𝘧𝘵𝘪𝘯𝘨 𝘰𝘧𝘧-𝘴𝘤𝘳𝘪𝘱𝘵) After some time of tinkering, I had something I’m happy with. Parlant addresses one of the main problems when working with LLMs, 𝗵𝗼𝘄 𝘁𝗼 𝗺𝗮𝗸𝗲 𝘁𝗵𝗲𝗺 𝗳𝗼𝗹𝗹𝗼𝘄 𝗮 “𝗹𝗼𝗴𝗶𝗰𝗮𝗹 𝗽𝗮𝘁𝗵” 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗵𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗻𝗴 𝗼𝗿 𝘀𝘂𝗴𝗴𝗲𝘀𝘁𝗶𝗻𝗴 𝗶𝗿𝗿𝗲𝗹𝗲𝘃𝗮𝗻𝘁 𝘁𝗼𝗽𝗶𝗰𝘀. What clicked for me: 1. 𝗙𝗹𝗼𝘄𝘀 𝗮𝗿𝗲 𝗲𝘅𝗽𝗹𝗶𝗰𝗶𝘁, so I can see exactly what the agent is allowed to do. 2. 𝗚𝘂𝗶𝗱𝗲𝗹𝗶𝗻𝗲𝘀 𝗹𝗶𝘃𝗲 𝗻𝗲𝘅𝘁 𝘁𝗼 𝘁𝗵𝗲 𝗰𝗼𝗱𝗲, which makes “don’t say X / always do Y” practical. 3. It’s 𝗲𝗮𝘀𝘆 𝘁𝗼 𝗿𝗲𝗽𝘂𝗿𝗽𝗼𝘀𝗲 the same setup for support, content, or simple sales workflows. 4. 𝗘𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴 𝗶𝘀 𝘀𝗼 𝘀𝗺𝗼𝗼𝘁𝗵! I've recorded a tutorial where I go from 𝗽𝗶𝗽 𝗶𝗻𝘀𝘁𝗮𝗹𝗹 → 𝗱𝗲𝗳𝗶𝗻𝗲 𝘁𝗼𝗼𝗹𝘀 & 𝗳𝗹𝗼𝘄𝘀 → 𝘁𝗲𝘀𝘁 𝗮 𝘄𝗼𝗿𝗸𝗶𝗻𝗴 𝗮𝗴𝗲𝗻𝘁 If you’re building in CX/content/data and want predictability over “vibes,” this might be useful. Go check my repo to get the full code 👩🏻💻 👉🏻 https://lnkd.in/di78RsJs Go check my full tutorial using it ✨ 👉🏻 https://lnkd.in/dvpPKxDY #AI #Parlant #CX #genai #llm #conversationalai #opensource

🚀 From customer idea → live in the Citrine Platform (in under a month!) We heard a clear request from a customer: “We’d like to try different machine learning algorithms in the UI, not just the API—so product experts can do it easily.” Our Customer Service Manager and Solution Engineer relayed the feedback to Product, and… we got to work. 🤝 ☑️ In the Citrine Platform UI, you can now tick a box to try alternative algorithms—no code needed. 🎯 The system automatically selects the most accurate algorithm for your problem. ⚡ Faster experimentation, broader access for product experts, and better results—without leaving the interface. Huge shout-out to our CS, SE, and Product teams for moving quickly and to our customers for the nudge. Keep the feedback coming—we’re listening, and we’re shipping. 💛 #ProductUpdate #CustomerObsessed #MachineLearning #NoCode #DataScience #CitrinePlatform #MaterialsInformatics

🚀 How to Chain Custom Agents in VS Code + GitHub Copilot — The Vision That Changes Everything ⸻ Ever imagined GitHub Copilot doing more than autocomplete — like actually orchestrating a mini AI dev team inside your VS Code workspace? That’s what I explored recently: chaining custom Copilot agents. ⸻ 🧠 The Concept Each agent has a specific role and behavior defined inside .github/chatmodes/ as a .chatmode.md file. Now imagine connecting them together so one’s output becomes the next one’s input — forming a multi-agent workflow. Here’s a simple chain: 🧩 DocAgent → 🧪 TestAgent → ⚙️ CIConfigurator • DocAgent — scans code and writes documentation • TestAgent — reads docs metadata, creates tests • CIConfigurator — builds CI pipeline to run those tests ⸻ 💡 Why This Is Exciting This moves Copilot from an “assistant” to a true AI teammate. Each repo can ship with its own .chatmodes folder, defining team-wide AI behavior — so your Copilot becomes project-aware. ⸻ In my next post, I’ll share the actual steps and file structure you can use to implement this chaining model — including .chatmode.md examples and orchestration logic 💻 Stay tuned 🔥 #GitHubCopilot #AI #VSCode #Developers #Automation #AIagents #OpenSource

🚀 From Data Chaos to API Harmony A few years back, our team was drowning in complexity. Every system spoke its own language, every feature had its own data format — and integrating them felt like solving a jigsaw puzzle with missing pieces. We realized something powerful: The problem wasn’t in our code, it was in our approach. So, we flipped the script. Instead of designing apps around UI or features first, we started building around data — clean, consistent, reusable data. That’s when the real transformation began. 📊 The Data-Centric Approach taught us that every service, every insight, every decision starts with how well we manage and understand our data. Once our data foundation became strong — the rest followed naturally. Next, we embraced API-First Development. APIs became the universal language between our systems — intuitive, scalable, and developer-friendly. Now, teams don’t wait for backends to finish. They design, mock, and test APIs from day one. And suddenly… ✨ Features shipped faster. ✨ Integration became simpler. ✨ Innovation picked up speed. Today, when someone asks how we scaled so seamlessly — my answer is simple: We stopped building systems. We started building data-driven ecosystems. #Architecture #API #DataDriven #TechLeadership #DigitalTransformation #Microservices #LovelyOnTech

Had a great time today presenting our latest findings at LeadDev New York. We published updated data on the amount of Copy/Paste (Duplicated) vs Moved (Refactored) code. When we evaluated this data just two quarters ago, we found 12% of code being copy/pasted vs. 9.5% move/refactor. Now, we observe 16.9% copy/paste and 6.9% move/refactor. 😬 It's hard to reconcile how a company can generate 30, 50 or 90% of their code via AI, yet still have such incremental improvement in productivity metrics like Commit Count (9% median increase since 2022, per our data & per Sundar Pichai) or Diff Delta (15%). With more direct AI usage measurement launching every week this month (we're already integrated with Cursor & Github Copilot), we're plan to have a lot more juicy data to share in the months to come. If you saw the talk, drop a comment, maybe we can catch up and chat about what you've observed?

“If you measure too much, you can’t see anything.” The art of observability isn’t about collecting more data — it’s about finding signal in the noise. True observability means: 🔹 Measuring what truly matters 🔹 Understanding what those metrics actually mean 🔹 Acting on insights before issues become incidents In a world drowning in dashboards and alerts, clarity is the new superpower. #Observability #DevOps #SRE #PlatformEngineering #Monitoring #DataDriven #Kubernetes #SiteReliability

Jedify recently launched Semantic Statements to turn everyday business language into consistent answers at scale. Your org already knows its business lingo: acronyms, product code names, customer labels. That knowledge used to live in docs or in people’s heads - now it lives in Jedify, so your data agent understands your domain and speaks your terms. Map common terms and synonyms, plus specifics like time rules and formatting, to see how Jedify interprets questions and composes answers in your language. Click below to read more about this new feature and watch a short demo.