Inside the Largest Agentic AI Conference for Developers

What happens when 2,000+ developers, 60 speakers, and one buzzword called 'agentic AI' walk into a convention centre in Bengaluru?‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  

Inside the Largest Agentic AI Conference for Developers THE BELAMY Weekly Newsletter of AIM Monday, March 30, 2026 | By Mohit Pandey Now, subscribe to our Digital & Print Editions > What happens when 2,000+ developers, 60 speakers, and one buzzword called 'agentic AI' walk into a convention centre in Bengaluru? They try to make AI act, decide, remember, and sometimes, break in production. That was MLDS 2026. Over two days at the NIMHANS Convention Centre, the Machine Learning Developers Summit, or MLDS 2026, moved past the usual model-centric conversations. This year was about agents. Not prompts, not models, but systems that can act on their own.  Even before entering the halls, developers were immersed in activity. The venue was lined with companies and their booths showcasing what they are building. With BITS Pilani Digital, Neysa, E2E Networks, Genpact, StoneX, Snowflake, Atlassian, Millennium Management, Intuit, Acceldata, EY, Orkes, MathCo, TiDB, Walmart Global Tech, and SonarQube, the corridors were packed. People sipped coffee, played chess, and discussed how they are 'vibe coding' in their free time. Click here to join the AIM Chess League. If there was one clear takeaway, it was this: the industry has stopped obsessing over models and started worrying about what happens after deployment. Across both days, the sessions circled back to the same problem. Building agents is easy; running them in production is not.  And the energy was simply crazy! The summit started off with Professor Snehanshu Saha, Head of AI Research (APPCAIR) at BITS Pilani, who spoke about real agentic AI use cases in delivery systems. This was followed by Kesava Reddy from E2E Networks, who demonstrated a 13-GPU fleet, showing what real infrastructure looks like beyond demos.  Sachin Tripathi and Abhishek Kumar from AIM led some of the most packed sessions at the summit. One focused on taking agentic systems from prototype to production, another explored what was described as "protocol wars", and a third examined how to make agents observable at scale. By the end, it was clear that the battle now is not just about models, but how agents talk to each other. MCP, A2A, context layers—everyone is building their own stack, and no one agrees on standards yet. This means the next bottleneck is not intelligence, but coordination. The first day ended with a special standup by comedian Anirban Dasgupta, who playfully engaged developers in the audience about what they are building, highlighting differences between the older generation of builders and today's developers. The Enterprise Reality Check A significant portion of MLDS 2026 was grounded in enterprise use cases. The message was how companies are experimenting heavily, but very few have figured out how to scale AI reliably. From Genpact to EY to Atlassian, speakers broke down what actually breaks when building agentic AI systems: observability, memory, evaluation and governance. Meanwhile, Sumeet Tandure and Sarita Priyadarshini from Snowflake focused on architecture. Multi-agent systems are quickly becoming the default, but designing them remains more art than science. Patterns are emerging, but there is no standard playbook yet. Even databases came under scrutiny. Arpit Bhayani from DiceDB presented a talk titled 'Databases Were Not Designed For This', which resonated strongly. Traditional data layers are struggling to keep up with dynamic, context-heavy, and constantly evolving agent workflows. Memory is the new frontier. If 2025 was about context windows, 2026 is about memory. Multiple sessions drilled into this—persistent memory, contextual recall, and long-running agents. RPTech, an NVIDIA Partner in association with AIM, to Host 'GenAI: From Build to Impact' Meetup in Hyderabad, showcasing DGX Spark for AI developers Rashi Peripherals Limited (RPTech), an NVIDIA partner, in association with AIM, is hosting GenAI: From Build to Impact, a developer meetup focused on building and scaling real-world generative AI applications, using its DGX Spark™ platform.  The event will showcase how developers can use NVIDIA DGX Spark™, the company's desk-side personal AI supercomputer, to accelerate the generative AI development lifecycle from model experimentation and fine-tuning to high-performance inference. Click here to find out more about GenAI: From Build to Impact. India's AI Stack Takes Shape One of the more interesting threads this year was the move beyond software. Sushovan Mukherjee, Associate Director at EY GDS Consulting and Abhijit Jana, Senior Director of Engineering and Robotics at West Pharmaceutical Services, touched on physical AI—robots, satellites, and embodied systems. Radha Krishna Kavuluru from Dhruva Space further carried the discussion into the space economy, addressing autonomous satellites and Earth intelligence systems. This is where agentic AI starts colliding with physics. Latency matters, and so does energy. Failure has real consequences. It also makes the current obsession with chat interfaces feel limited. A strong India narrative ran through the summit. Ganesh Ramakrishnan's BharatGen talk stood out. The focus was clear. Multilingual, multimodal AI built for India, at scale, and at low cost. The company's flagship model, Param2, was a highlight. Alok Sharma from Meesho highlighted in his session on multilingual voicebots for Bharat. The next wave of AI adoption in India will not be English-first. It will be voice-driven, regional, and deeply contextual. EkStep's Shalini Kapoor's session, 'AI Ready Data for India', added another perspective. Data infrastructure in India is still fragmented. Without fixing this, scaling agentic systems will remain difficult. The Developer Mood and Cheteshwar Pujara What MLDS does well is capture the developer sentiment early. And this year, the mood was a mix of excitement and fatigue. Excitement, because agentic AI feels like a real shift. Systems that can act, not just respond. Fatigue, because the stack is becoming complicated with LLMOps, evaluation pipelines, memory layers, orchestration frameworks, and protocol standards. Everyone is building. No one feels finished. There is also a growing honesty in the ecosystem. Fewer flashy demos. More conversations about failure, cost, and tradeoffs. But overall, the sentiment towards building AI in India and for India remains consistent. To close on a note that is close to everyone's heart, MLDS 2026 ended with cricket. And then came Cheteshwar Pujara on stage. In a room full of developers talking about latency and throughput, a cricketer talking about patience could have felt out of place. But it didn't. His session, titled 'Still Here—The Long Game', landed harder than expected. Pujara spoke about staying relevant in a system constantly chasing the next big thing, about discipline, consistency, and doing the boring work repeatedly. It mapped almost perfectly to where AI is today. 40 Under 40 AI Builders The 40 Under 40 AI Builders Awards followed, presented by Pujara. This is where MLDS shifts from ideas to people. The award recognised people making visible and meaningful contributions to the AI ecosystem in India across startups, enterprises and GCCs. MLDS 2026 was not about breakthroughs. There was no single moment that changed everything. Instead, it felt like the industry was maturing.  If MLDS 2025 was about discovering agents, MLDS 2026 was about surviving them in production. The next year will decide who actually makes them work. AIM March Print Edition >> The March 2026 edition of AIM Print captures India's AI ecosystem without filtering. With BITS Pilani Digital redefining education under Rajiv Tandon, Anthropic CEO Dario Amodei positioning India as the most important market, and voice AI taking shape across several startups, the edition captures the shift that the country is undergoing with agentic AI. Click here to subscribe to AIM's Print. Now, subscribe to our Digital & Print Editions > For Brand collaborations, reply to this email or write to info@aim.media You received this email because you signed up to the updates from AIM. Click here to unsubscribe if you do not want to receive emails from us.