← Back to Speaking

Ship Sabha — Speaking on AI Evaluations

Arjun Thakur speaking on AI Evaluations at Ship Sabha, hosted by Amura Health, Chennai
Presenting at Ship Sabha — Chennai, 4 July 2026.

I spoke at Ship Sabha, a full-day gathering hosted by Amura Health in Chennai, on how organizations can evaluate, optimize, and scale AI systems for real-world impact. The theme of this edition was AI Evaluations — the discipline that separates AI demos from AI you can put in front of paying customers.

Event at a glance

  • Event: Ship Sabha — AI Evaluations edition
  • Host: Amura Health
  • Date: Saturday, 4 July 2026
  • Time: 9:00 AM – 6:00 PM (Morning: Learn · Afternoon: Ship)
  • Venue: Amura Health, One National Park, Velachery, Chennai, Tamil Nadu, India
  • Format: Talks + hands-on building — founders, CTOs, AI engineers, researchers, and students
  • Event page: luma.com/2m88m8l4

Why this event

Ship Sabha brings together founders, technologists, researchers, industry leaders, and builders to explore the ideas and challenges shaping the AI era. What makes it different from a typical meetup is the mandate: every edition aims to ship something tangible — an initiative, a project, a contribution that outlives the event. Morning is for learning; the afternoon is for building. That bias toward output is exactly the kind of room I like to speak in.

What I spoke about

My session focused on evaluations as the backbone of production AI — drawn directly from building Yuvan, a voice-first AI tutor now used by paying schools. The through-line: you don't earn trust in an AI system with a good demo, you earn it with evals, traces, and guardrails that hold up under real usage.

  • Why evals come first: treating evaluation as a first-class part of the system, not an afterthought bolted on before launch.
  • Trace-level observability: per-node traces and dataset-driven regression runs so every prompt change is measurable and replayable, not a guess.
  • Guardrails and verifiers: constraining reasoning to retrieved, grounded context and re-checking claims before the system speaks — the model proposes, the guardrails dispose.
  • Scaling economics: retrieval precision, semantic caching, and cost engineering so quality holds as usage grows.

The speaker lineup

It was a strong room to share the stage with:

  • Sudharsan D — Head of Engineering, iamneo
  • Lakshminarayanan Ramakrishnan — Co-Founder & CTO, Turiyatree Technologies
  • Arjun Thakur — Founder & Principal AI Engineer, Yuvan

Moments from the day

Panel discussion at Ship Sabha with Arjun Thakur and fellow speakers, moderated on stage
Founders, CTOs and engineers in the audience at Ship Sabha, Chennai

Bring a session like this to your team or event

I speak on agentic AI, evaluations, RAG, and voice AI — for conferences, engineering teams, and universities, in India and abroad. Sessions are practitioner-first: live code, real numbers from systems in production, and takeaways your team can act on the next morning. If you're organizing an event or want to upskill an engineering org, let's talk.