About
With over 6 years of experience in data and artificial intelligence, I specialize in…
Activity
3K followers
-
Alan-Michael Khalil reposted thisAlan-Michael Khalil reposted thisAlan-Michael Khalil just wrapped up an insightful session at Code Europe 2026 on orchestrating AI agent teams for marketing outcomes. 👏 Thank you to everyone who joined us to see how HelloFresh Group leverages intelligent automation to scale our global AdTech capabilities. Key takeaways from his presentation: 🟢 Scalable AI Orchestration: Coordinating specialized agent teams to execute complex, multi-step marketing workflows efficiently. 🟢 Measurable Impact: Moving beyond simple AI integrations to deploy agents that drive targeted personalization and high-performance campaigns. 🟢 Engineering Infrastructure: The backend architecture required to sustain and optimize AI-driven marketing strategies at a global scale. If you missed the session or want to dive deeper into our AdTech ecosystem, connect with Alan-Michael or share your thoughts in the comments. 💡 We are always looking for passionate tech builders to help us change the way people eat forever! #CodeEurope #CodeEurope2026 #HelloFresh #HelloFreshTech #AdTech #AIEngineering #TechBuilders
-
Alan-Michael Khalil shared thisI will be speaking virtually at next week's Code Europe conference in Poland. My talk will cover how we orchestrate AI agent teams for marketing outcomes at HelloFresh. I’ll share some of the real, agent-native workflows creating incredible leverage for our marketing organization, as well as the less glamorous engineering challenges involved in making them reliable in production. https://lnkd.in/gDNNn74gHow Hellofresh Group orchestrates agent teams for marketing outcomes — Code Europe 2026How Hellofresh Group orchestrates agent teams for marketing outcomes — Code Europe 2026
-
Alan-Michael Khalil reposted thisAlan-Michael Khalil reposted thisWe’re growing our Data Science + ML teams at HelloFresh 🚀 If you’re interested in making an impact at scale, these are worth a look. This work sits within our Growth Alliance, where experimentation is at the core of how we drive decisions, optimize performance, and continuously improve the customer experience. That’s exactly where these roles come in: 👉 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿 👉 𝗦𝗲𝗻𝗶𝗼𝗿 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿 👉 𝗦𝗲𝗻𝗶𝗼𝗿 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁 If this is the kind of environment you’ve been looking for, definitely take a look 👀 Links are in the comments!
-
Alan-Michael Khalil shared thisA high-impact Sr MLE role just opened up on a colleagues team. The position is focused on scaling MLOps and ML Solutions across our global company. Learn more about the role in the link - happy to chat if you'd like a referral. Benjamin Bertincourt Michael Johnson https://lnkd.in/eF7sNZARSenior Machine Learning Engineer [Decisions] in Toronto, Ontario, Canada | Software Engineering at HelloFreshSenior Machine Learning Engineer [Decisions] in Toronto, Ontario, Canada | Software Engineering at HelloFresh
-
Alan-Michael Khalil shared this📣 I'm excited to be speaking at the upcoming CDAO Canada 2025 Summit! Join me as we dive into what's now and next in predictive and generative AI in industry. 📅 Date:March 26-27 📍 Location: Hyatt Regency Toronto. 🎫 To join me, get your pass here: https://lnkd.in/e-P2SxWU #CDAOCanada #dataanalytics #data #analytics #CDAO #CDO #dataevent Corinium Global Intelligence
-
Alan-Michael Khalil posted thisIf your AI/ML platform is 𝗺𝗼𝗱𝘂𝗹𝗮𝗿 and 𝗮𝗱𝗮𝗽𝘁𝗶𝘃𝗲, your personal stock price is up today. AI tech (the entire tech stack/sandwich) is evolving at lightning speed. Organizations that keep up do two things really well: 1. They build modular systems. This means you can add new features without breaking the whole design. It’s like upgrading your phone without throwing out your apps. 2. They stay adaptive. Think of it like being able to shift gears when the road changes—adjusting quickly to new use cases, customer needs, or even surprise opportunities and challenges. In 2025, being flexible isn’t optional—it’s a game-changer. #deepseek #deepseekR1 #openAIoN #gemini2.x
-
Alan-Michael Khalil shared this𝗧𝗵𝗲 𝘀𝘁𝗲𝗲𝗽 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗰𝘂𝗿𝘃𝗲 𝗳𝗼𝗿 𝗰𝗼𝗻𝘁𝗮𝗰𝘁 𝗰𝗲𝗻𝘁𝗲𝗿 𝗮𝗴𝗲𝗻𝘁𝘀 𝗶𝘀𝗻’𝘁 𝗷𝘂𝘀𝘁 𝘁𝗼𝘂𝗴𝗵 𝗼𝗻 𝗲𝗺𝗽𝗹𝗼𝘆𝗲𝗲𝘀—𝗶𝘁’𝘀 𝘁𝗼𝘂𝗴𝗵 𝗼𝗻 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿𝘀 𝘁𝗼𝗼. (Have you ever experienced a frustrating call with customer support? This might be why.) Traditional training methods often leave agents unprepared for the complexities of real-world interactions. The result? Uncertainty and frustration on both ends of the line. At Bell, we asked: 𝗪𝗵𝗮𝘁 𝗶𝗳 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗰𝗼𝘂𝗹𝗱 𝗯𝗲 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁? What if new hires could navigate scenarios like assisting a frustrated customer or a complex billing issue before handling live calls? We completely transformed this experience at scale with an AI-driven solution that’s redefining agent training. Agents can practice handling diverse customer personalities in a safe, realistic environment, while receiving personalized feedback. The results? • Shorter onboarding times • Improved agent performance • Reduced turnover • Better customer experiences This isn’t just training; 𝗶𝘁’𝘀 𝗰𝗼𝗻𝗳𝗶𝗱𝗲𝗻𝗰𝗲-𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 ��𝘁 𝘀𝗰𝗮𝗹𝗲. 🔗 Curious to learn more? Check it out here: https://lnkd.in/eUxni_94Empower your agents with the Bell Customer Simulator, powered by AIEmpower your agents with the Bell Customer Simulator, powered by AI
-
Alan-Michael Khalil shared thisUnlock scalability, operational efficiency, and AI potential 🚀 from any cloud data migration project. Check out these six key steps for a seamless migration to the cloud 🌐Alan-Michael Khalil shared thisDid you know that #cloud migrations can supercharge business operations, including #AI capabilities? At Bell, we saw a remarkable 30% increase in AI model performance and a 95% reduction in processing time for some workloads after migrating our data to the cloud. Discover how to achieve similar success by reading these 6 essential tips for cloud migration from our Director of Data Engineering, Ryan Levman: https://lnkd.in/ehTTSUFHSix key steps for a successful cloud data migrationSix key steps for a successful cloud data migration
Alan-Michael Khalil commented on a post
3w
Congrats on shipping Salma Bezrati
-
Alan-Michael Khalil liked thisAlan-Michael Khalil liked thisInteresting launch from Databricks: ai_decide(). A few weeks ago, TypeSafe AI introduced Jev, built around a simple idea: not every AI task needs an LLM generating text. Sometimes software just needs a fast, structured decision. Databricks is now bringing that idea closer to enterprise data with ai_decide(). Instead of asking an LLM to generate an answer, you can directly make decisions such as: * Which model should handle this prompt? * Does this document need human review? * Which category does this request belong to? * Did an AI agent follow the expected policy? * What action should an application take next? And because it can run directly against governed Databricks data through SQL or REST, this gets very interesting for large-scale data pipelines and AI agents. I think we’ll see more of this pattern: LLMs for reasoning and generation. Decision models like Jev for fast, cheap, structured decisions. #Databricks #Jev #AI #DataEngineering #AIAgents #GenAI
-
Alan-Michael Khalil liked thisAlan-Michael Khalil liked this📦 First box. Second box. Weekly habit. If you’re a product leader who loves figuring out what makes customers come back, I’m hiring. We’re looking for a #StaffProductManager at HelloFresh in #NewYork or #Toronto to lead our #Onboarding squad. The opportunity? Own the journey from first delivery to lasting loyalty across HelloFresh Group brands, including Factor_. Here’s what makes this role exciting: 🚀 Build from 0→1: Shape an ML-powered “Next Best Action” system that helps each customer take the right next step. 🌍 Work across brands and channels: Bring together web, app, and lifecycle communications into one connected experience. 🧪 Turn insights into impact: Run experiments that improve activation and early retention. 🤝 Lead across teams: Partner with Engineering, Design, Data Science, CRM, and Marketing to bring the vision to life. You’ll bring 6+ years of product management experience, strong consumer subscription and growth expertise, and the drive to turn ambitious ideas into shipped products. Sound like your next challenge? Explore the role and apply here. Know someone who would crush this? Tag them below or share this post. 👇 #Hiring #ProductManagement #Growth #Onboarding #HelloFresh
-
Alan-Michael Khalil liked thisAlan-Michael Khalil liked thisI first worked on search indexes almost 20 years ago, building full-text search in SQL Server. Twenty years later, I'm still working on search in the database. Why? 1. There's a new consumer running search. Retrieval for Agents brings a new set of requirements for scale, qps, and cost. 2. There's a new architecture for databases. Lakebase gives us a foundation to run operational workloads with serverless + object-store economics. So we built new search extensions for Lakebase Postgres, and the results are epic: - Fastest and most efficient system on VectorDBBench 100M. - Runs 4x more cost-effectively than standard pgvector setups. - Powers hybrid search with native BM25. - Scales entirely with usage (incl. scale to 0). Lakebase Search was shaped by feedback from hundreds of customers, and today it's GA. Search in a OLTP database is no longer a compromise. https://lnkd.in/eCBwvUiQLakebase Search: State-of-the-art full text and vector search for PostgresLakebase Search: State-of-the-art full text and vector search for Postgres
-
Alan-Michael Khalil liked thisAlan-Michael Khalil liked thisLast week, we had the first Data and AI summit at HelloFresh, jointly organized by HelloFresh and Databricks. One major topic that a couple of talks circled around is what is necessary for Agentic AI to succeed in enterprises. Model intelligence is solved for most tasks in enterprises. But Agentic AI tends to fail on four other crucial fronts, captured by the 4C framework: 🧠 Context: Agents need the Enterprises’ data in an easy-to-process format to avoid the agent going on a costly and slow random walk through the data. That context layer is provided by Genie Ontology. 🔀 Choice: No single foundational model stays best (=most cost-efficient) for long. Enterprises need to stay flexible. With Model Serving and Unity Gateway you swap and route across open, proprietary and fine-tuned models. 🛡️ Control: All Agents in production need to be governed, lineage, evaluation, guardrails, and cost controls are crucial. Unity Gateway provides this unified for all Agents. 💰 Cost: obviously super important; only if Context, Choice, and Control are set up are the right tools for monitoring and optimizing Cost available.
-
Alan-Michael Khalil liked thisThis will be interesting!Alan-Michael Khalil liked this𝗖𝗮𝗻 𝗝𝗲𝘃 𝗿𝗲𝗽𝗹𝗮𝗰𝗲 𝘆𝗼𝘂𝗿 𝗟𝗟𝗠 𝗷𝘂𝗱𝗴𝗲? ⚖️ We compared Jev, TypeSafe’s model for structured decisions, with GPT, Claude, and DeepSeek using MLflow. On 30 human-labeled QA examples, Jev achieved: ✅ 𝟯𝟬/𝟯𝟬 agreement with human labels ⚡ 𝟯𝟲𝟵 𝗺𝘀 median latency 💰 $𝟬.𝟬𝟮𝟰𝟳 estimated cost per 1,000 judgments Jev matched the GPT judges’ agreement with human labels and had the lowest measured latency and estimated cost in this test. One tradeoff though is that Jev cannot provide a written rationale explaining its verdict. Our new blog walks through building a Jev scorer in MLflow, measuring quality, cost, and latency, and deciding when to use another judge. Try the same comparison on your own dataset. 🔗 𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗳𝘂𝗹𝗹 𝗰𝗼𝗺𝗽𝗮𝗿𝗶𝘀𝗼𝗻: link in comments. 👇 #MLflow #LLMOps #GenAI #AIEngineering
-
Alan-Michael Khalil liked thisAlan-Michael Khalil liked thisThe author copies of my first book arrived today! After months of writing, reviewing, rewriting, testing code, debating details, and somehow getting everything across the finish line, Practical MLflow for Generative AI on Databricks is now officially in print. I’m incredibly grateful to have worked on this with my co-authors Julie Nguyen and Lim Chang Shi. This book is very much a product of our collective experience, discussions, and many rounds of refinement. A special thanks to Matei Zaharia for writing the foreword and Max Thöne for the generous back-cover endorsement. I also want to thank all of our technical reviewers who spent their time challenging our ideas, catching issues, and helping us make the content more useful and technically rigorous. And a huge thank you to the O'Reilly editorial, design, and production teams who guided us throughout the process and turned hundreds of pages of drafts, figures, and code into the book I’m holding today. The book follows the lifecycle of building production AI agents using MLflow and Databricks, from prompt development and tracing through evaluation, agents, deployment, and production monitoring. If you work with #GenAI, #AIEngineering, #LLMOps, #MLOps, #MLflow, or #Databricks, I hope you find something useful in it. And if you do read it, I would genuinely love to hear your feedback. 📖 Amazon: https://lnkd.in/gBZxyZAq 💻 Companion code: https://lnkd.in/dBKhbUBk
-
Alan-Michael Khalil liked thisAlan-Michael Khalil liked thisRound two at Iterable Activate New York and the theme is getting clearer: AI is only as good as the data, and the judgment, behind it. 🧠 AI in Practice: Three Marketing Teams. Three Lessons. Marketers from Trek Bicycle, Square, and Babylist shared how they're putting MCP to work, and it was refreshingly practical. We all know the swivel-chair moment of copying and pasting between tools. MCP is the solve. Every manual cleanup step is a tax that compounds, and so much of it (auditing included) could run through an agent instead. But the reminder I needed: AI doesn't know your marketing. You do. The best framing of the day was by Jenn Clauss: the agent doesn't decide, it classifies. Judgment on risk stays with a human. And before you build anything, ask: "Where does the answer live?" Quickly shifted over to where attention lies with a brilliant panel. The Washington Post, HelloFresh, and Twilio unpacked what it takes to earn attention now. Customers expect to be known because they know you have the data. One size fits all is no longer an option. That means a seamless, consistent experience across email, app, text, and web, and consistency across departments too. None of it works without clean data. That part is non-negotiable. Ami Upton wrapped up the session with three key takeaways: Attention follows trust. Relevance beats volume. Simple wins. #IterableActivate #AIinMarketing #MCP #CustomerExperience #MarketingLeadership
-
Alan-Michael Khalil liked thisAlan-Michael Khalil liked thisAs I wrap up my time at HelloFresh, I am blown away with the impact this opportunity has allowed me to have. Over the past 10 weeks, I got to build AI tools and internal dashboards that real people at the company use. This experience is a different kind of learning than anything a classroom gives you. You figure out fast what matters and what doesn't when something you built is in front of someone trying to do their job. A few things I'm taking with me: 1) How to move fast without breaking trust 2) That the best tools solve a real problem, not a hypothetical one 3) That AI is most powerful when it's invisible and just makes someone's day easier Grateful for Brian Leung and Jenna Kirschner for teaching me the ropes, hopefully I have passed executive communication 101. Keep your eyes Peeled for what's next. 🙌 👀 #HelloFresh #Internship #AI #ProductDesign
-
Alan-Michael Khalil liked thisAlan-Michael Khalil liked thisJoin us at AI Infra Nights x Autonomous Engineering and catch Dhruv Gupta’s session, Omnigent: A Meta-Harness for AI Agents. 🙌 He’ll walk through orchestrating multiple coding agent harnesses, keeping control in a vibe-coding workflow, and a demo of governed multi-harness setups. Also on the agenda: talks from Islo, Alien, and Headroom Labs. 🗓️ Thurs, Sept 24 🕜 6:00 PM PT 📍 Databricks HQ, SF Space is limited. Register: https://luma.com/brebu7ww #Omnigent #AIAgents #OpenSource
Experience & Education
Licenses & Certifications
Volunteer Experience
Courses
-
SAS Macro Language 1: Essentials
-
-
SAS Programming 1: Essentials
-
-
Statistical Methods for Business
-
-
Statistical Modelling
-
Other similar profiles
Explore more posts
-
Vishen Maharaj
Nearly 90% of Canada’s mid‑market leaders say they’re confident in their AI progress — and most are already seeing ROI. But confidence isn’t the same as maturity. Our latest report, The Business of AI 2026 (MNP Digital), shows that while AI is firmly operational, only a small minority are using it systemically or as a true strategic differentiator. The gap isn’t technology. It’s skills, governance, and execution discipline. AI advantage won’t come from more tools — it will come from sharper choices about where AI drives outcomes that actually matter. #AI #ExecutivePerspective #DigitalStrategy #ArtificialIntelligence #MidMarket #MNPDigital
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More