Call for proposals: Fifthel Hyderabad meet-up

28 November 2026 · Proposal deadline: 31 October 2026

Overview

We welcome talks about work that shipped to production, and work that failed to ship, for the right reasons.

This meet-up is about people sharing their stories: the work, the architecture, the planning, the iterations before landing in production, the aftershocks, the bug fixes and hot patches that followed, and the triaging that found the root cause.

A proposal does not need to describe a finished or polished system. A work-in-progress with real findings beats a finished demo with none.

Where possible, include an architecture walkthrough, a working example, or a demo that helps explain your findings.

With that in mind, here are the three topics we’re interested in for this round.

Topics of interest

Agents in production

The core focus for this event is agents that are live and taking real traffic: chat assistants, coding agents, voice bots, workflow automation wired into CRM, ticketing or internal tools, or anything else in that shape. We’re not fixed on the form the agent takes; we’re interested in what happened after it went live.

We also welcome work that stopped short of production when you can explain what you learned and why you chose not to ship.

If you’re not sure what to cover, start here:

• What did going from prototype to production actually cost—in latency, dollars, and engineering time?
• Where did the agent fail in a way your evals didn’t catch, and how did you find out?
• What’s the recovery path when the agent gets it wrong—retry, escalation to a human, or silent failure?
• What did you have to build around the agent—guardrails, monitoring, fallback logic—that you hadn’t planned for?

Applied AI

RAG, MCP, skills, plugins, and agent tooling: how these get built into a product that ships. Stories about integration pain, cost, and what the first version got wrong are as welcome as the polished result.

If you’re not sure what to cover, start here:

• What did you try first that didn’t work, and what made you change your approach?
• Where did the abstraction leak? What did the framework or library not tell you until it hit production?
• What did you end up building yourself because the off-the-shelf piece didn’t fit?
• What surprised you once real data and real users hit it?
• Have you turned tools like Claude Code, Codex, or custom agent harnesses into shared team infrastructure through skills, plugins, marketplaces, or reusable workflows? What changed when colleagues started using and extending them?

Fine-tuning, RL environments, and small language models

We welcome experiences with fine-tuning runs that did or didn’t beat prompting, distillation and quantization, reward design, RL environments and their harnesses, and small models running on-device or at the edge.

Much of the best work here sits behind an NDA. If you can’t share exact numbers or setup, walk through an open-source equivalent, or explain the architecture and techniques without proprietary specifics. The shape of the work is what we’re after.

If you’re not sure what to cover, start here:

• What made you fine-tune instead of prompt, and did that decision hold up?
• How did you build the environment and the reward, and what did the model learn to game?
• What did the smaller model cost you in quality, and what did it buy back in latency or spend?
• What does the retraining loop look like now that it’s running?
• What have you built with local or on-device models—from the Qwen, Gemma, Llama, or other model families—for voice, classification, vision and video, privacy, or another specialized use case? What did you have to build or change around the model to make it work?

Session formats

• 45 minutes: talks that include demos.
• 30 minutes: talks without demos.
• 15 minutes: flash talks or quick demo-based talks.

Please indicate your preferred format in your proposal.

How to submit

Submit your abstract by 31 October 2026.

Your abstract should clearly describe:

• The problem or topic being addressed.
• Why it is relevant.
• Who the session is intended for.
• The key takeaways attendees can expect.

About the editorial team

Madhusudhan Sambojhu — editorial lead. Co-founder & Lead Engineer at Able & Thoughtful Robots. An active reviewer in the Fifthel and Rootconf communities, Madhusudhan leads the editorial process for this edition.

Chaitanya Sangani — reviewer. An independent AI consultant and member of The Fifth Elephant community. Chaitanya was the editor for the September Hyderabad meet-up and has reviewed talks for the Pune edition and The Fifth Elephant 2026 Annual Conference.

Srix Sriramkumar — reviewer. Srix works with engineering teams to make them AI-native. A member of The Fifth Elephant community, he has reviewed submissions for the Enterprise AI in production meet-up, held in Bangalore in June.

Contact

Call/WhatsApp - (91)7676332020
Email - info@hasgeek.com
Comments - https://hasgeek.com/fifthelephant/hyd-meetup-cfp/comments