BEGIN:VCALENDAR VERSION:2.0 PRODID:-//HasGeek//NONSGML Funnel//EN DESCRIPTION:Share what it takes to run AI inside an enterprise: architectu res\, trade-offs\, and lessons from production. X-WR-CALDESC:Share what it takes to run AI inside an enterprise: architect ures\, trade-offs\, and lessons from production. NAME:Enterprise AI in Production: Mumbai Call for Proposals X-WR-CALNAME:Enterprise AI in Production: Mumbai Call for Proposals REFRESH-INTERVAL;VALUE=DURATION:PT12H SUMMARY:Enterprise AI in Production: Mumbai Call for Proposals TIMEZONE-ID:Asia/Kolkata X-PUBLISHED-TTL:PT12H X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT SUMMARY:Enterprise AI in Production: Mumbai Call for Proposals DTSTART:20261031T053000Z DTEND:20261031T103000Z DTSTAMP:20261002T231939Z UID:session/BcvEBiutsJvDkuem1dPS9t@hasgeek.com SEQUENCE:9 CREATED:20261002T033044Z DESCRIPTION:# Call for Proposals: Enterprise AI in Production\, Mumbai\n\n **Saturday\, 31 October 2026 · 11 AM–4 PM IST · Mumbai** \n**Deadline to submit by Sunday\, 11 October 2026\, 11:59 PM IST**\n\nVenue will be a nnounced by **7 October 2026**.\n\nLive streaming will be available for [F ifthel annual members](https://hasgeek.com/fifthelephant#memberships). All talks will be recorded.\n\n**Submit your proposal — https://hasgeek.com /fifthelephant/enterprise-ai-in-production-mumbai-cfp/sub**\n\n## Why this meet-up\nPutting AI into an enterprise means getting through security rev iews\, integrating with legacy systems\, working within budgets\, and earn ing the trust of the people who use it.\n\nThis meet-up brings together ar chitects\, platform engineers\, data leaders\, security and risk practitio ners\, and delivery teams to share what that work actually involves. Mumba i’s banks\, insurers\, payment networks\, and industrial and consumer en terprises provide a setting for these conversations.\n\nWe want architectu res\, numbers\, trade-offs\, and failure stories: what you put into produc tion\, what broke\, and what you changed.\n\n## Who should submit\nSubmit if you have worked on an enterprise AI deployment and can explain its impl ementation and outcomes. We welcome proposals from:\n\n- Enterprise and so lution architects designing integration patterns\, shared platforms\, and guardrails.\n- Platform\, ML\, and software engineers operating LLM\, retr ieval\, or agent workloads in production.\n- Data engineering and governan ce practitioners making enterprise data usable\, permissioned\, and audita ble.\n- Security\, risk\, compliance\, and audit practitioners who have ap proved\, blocked\, or reshaped deployments.\n- Delivery and programme lead ers who can connect implementation decisions to sustained adoption and mea surable results.\n\nWe especially welcome experience from banking\, financ ial services\, insurance\, payments\, healthcare\, telecom\, manufacturing \, and the public sector. First-time speakers are welcome\; prior speaking experience is not a requirement.\n\n## What we want to hear about\n\n### Reference architecture and platform design\nHow do you build an AI platfor m that multiple teams can use safely? Share decisions about central versus team ownership\, model gateways\, routing and fallbacks\, hosted versus s elf-hosted models\, and build versus buy. Explain the boundaries you chose and how they held up.\n\n### Integration with legacy and core systems\nSh ow how AI works with mainframes\, ERPs\, core banking systems\, and existi ng APIs. Topics include safe tool access\, MCP or API wrappers\, transacti on boundaries\, idempotency\, rollback\, and working within change windows and release calendars.\n\n### Data readiness\, retrieval\, and knowledge quality\nShare lessons from enterprise RAG\, document permissions\, freshn ess\, metadata\, deduplication\, and retrieval evaluation. We also welcome structured-data and text-to-SQL case studies\, including semantic layers and where they failed.\n\n### Governance\, risk\, and audit\nExplain how y ou turned policy and applicable regulatory requirements into working contr ols. Topics include approval workflows\, model inventories\, audit trails\ , data residency\, sensitive-data handling\, vendor risk\, and kill switch es. Be specific about the requirements and organisational context of your deployment.\n\n### Evaluation\, testing\, and reliability\nHow do you know the system is working? Show evaluation sets built from real workflows\, r egression testing across model and data changes\, tracing\, incident respo nse\, and lessons from production failures.\n\n### Agents and workflow aut omation\nWe want agents with real permissions and consequences. Share how you handle identity\, authorisation\, human approval\, prompt injection\, loops\, runaway costs\, and silent errors—and how you measure whether au tomation helps.\n\n### Cost\, capacity\, and unit economics\nExplain infer ence costs\, GPU capacity\, caching\, smaller-model strategies\, and cost attribution across teams. Share the production bill\, the assumptions that changed\, or why you decided to stop running a use case.\n\n### Adoption and operating models\nShow how team structures\, skills\, user feedback\, and operational ownership affected adoption. Connect organisational lesson s to a specific implementation\, including use cases that did not deliver the expected results.\n\nProposals may address several themes or explore o ne in depth.\n\n## Session formats\n\n| Format | Total session time\, incl uding questions | Best suited to |\n| --- | --- | --- |\n| Talk | 30 minut es | One system or decision\, with architecture\, trade-offs\, and outcome s |\n| Deep dive | 45 minutes | A complex implementation with diagrams\, d etailed examples\, or a walkthrough |\n\nChoose the format that fits your proposal. The editorial team may suggest a different format during review. \n\n## What makes a strong proposal\n\n- **Production evidence:** a system that is live\, or was live and was retired for reasons you can explain. I nclude duration in production and relevant measures such as users\, reques ts\, latency\, cost\, or error rates.\n- **Architecture you can explain:** components\, data flows\, trust boundaries\, and choices you rejected.\n- **Honest trade-offs:** what failed\, how you discovered it\, and what you changed.\n- **Clear constraints:** legacy systems\, security requirements \, budget\, headcount\, procurement\, or operating conditions.\n- **Useful takeaways:** specific lessons another practitioner can apply.\n\nWe will decline sales pitches\, tool tours without production context\, generic fr ameworks\, and principles without implementation evidence. Vendor practiti oners are welcome when the proposal shares substantive implementation less ons and clearly discloses commercial interests.\n\nIf your work is confide ntial\, anonymise the organisation and use ranges or relative measures whe re necessary. Explain these limitations in your proposal. Submit only mate rial you are authorised to share\; do not include confidential information in attachments.\n\n## What to include\n1. **Title and one-line pitch:** w hat will the audience learn?\n2. **Abstract\, 200–400 words:** the probl em\, implementation\, key decisions\, and outcome.\n3. **Architecture summ ary:** components\, data flows\, models\, and tools. A diagram is welcome. \n4. **Production evidence:** scale\, time in production\, measured result s\, and how you measured them. State where figures are approximate or anon ymised.\n5. **Failures and trade-offs:** at least one thing that went wron g or that you would do differently.\n6. **Three takeaways:** specific thin gs attendees can apply.\n7. **Audience and prerequisites:** who the sessio n is for and what they should already know.\n8. **Preferred format:** 30-m inute talk or 45-minute deep dive\, including questions.\n9. **Speaker det ails:** name\, role\, organisation\, short biography\, contact details\, a nd a professional profile. Prior talks or writing are optional.\n10. **Dis closures and availability:** commercial interests\, confidentiality constr aints\, and your availability to speak in Mumbai on 31 October and take pa rt in editorial preparation.\n\nDraft slides or a short recorded walkthrou gh are optional. \n\n## Selection and preparation\nThe editorial team will review proposals for relevance\, implementation depth\, evidence\, clarit y\, and usefulness to practitioners. We will also consider the balance of topics across the programme.\n\nShortlisted speakers may be asked for clar ifications or revisions. Accepted speakers should plan for an editorial re view of their slides before the meet-up.\n\n| Milestone | Date |\n| --- | --- |\n| Venue announced | By Wednesday\, 7 October |\n| Proposals close | Sunday\, 11 October\, 11:59 PM IST |\n| Clarifications and proposal revis ions | 12–14 October |\n| Selection decisions sent | Thursday\, 15 Octob er |\n| Selected speakers confirm participation | Friday\, 16 October |\n| Schedule announced | Monday\, 19 October |\n| First slide drafts due | Fr iday\, 23 October |\n| Editorial reviews and walkthroughs | 24–28 Octobe r |\n| Final slides due | Thursday\, 29 October |\n| Meet-up in Mumbai | S aturday\, 31 October 2026\, 11 AM–4 PM IST |\n\nAll deadlines are in Ind ian Standard Time (IST).\n\n## Submit your proposal\n\n**Submit your propo sal — https://hasgeek.com/fifthelephant/enterprise-ai-in-production-mumb ai-cfp/sub** by **11 October\, 11:59 PM IST**.\n\nFor questions or early f eedback on an idea\, contact info@hasgeek.com or 7676332020\n\nIf you have solved a difficult problem in putting AI to work inside an enterprise\, w e want to hear what you learned. LAST-MODIFIED:20261002T033044Z LOCATION:Mumbai - https://hasgeek.com/fifthelephant/enterprise-ai-in-produ ction-mumbai-cfp/ ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/enterprise-ai-in-production-mumbai-c fp/ BEGIN:VALARM ACTION:display DESCRIPTION:Enterprise AI in Production: Mumbai Call for Proposals in 5 mi nutes TRIGGER:-PT5M END:VALARM END:VEVENT END:VCALENDAR