Guests have been awestruck, audio system have been exhilarated, and the AI group was introduced collectively like by no means earlier than, as India’s most futuristic occasion concluded on Friday, August 22, 2025. Day 3 of DataHack Summit 2025, popularly often called DHS 2025, proved as soon as and for all that the one-of-a-kind occasion by AnalyticsVidhya wasn’t nearly talks and demos; it was about charting the course for the way forward for AI in India and past. From voice-driven interactions to no-code workflows. From securing AI techniques to scaling multi-agent collaboration. The summit’s final chapter left attendees impressed, knowledgeable, and keen to construct.
In case you missed the sooner days of DHS 2025, you’ll be able to learn all concerning the happenings of DHS 2025 Day 1 and Day 2 right here.
DHS 2025, Day 3: When GenAI Meets Actual-World Influence
The day started with Vijay Gabale, Co-Founder and CPO of Infilect, who delivered a stirring session on Why GenAI and LLMs Fail and How High-quality-Tuning Helps Them. With relatable, real-world examples, he reminded the viewers that even probably the most superior fashions stumble at fundamental logic. However with light-weight, domain-specific fine-tuning, these identical techniques can remodel into precision instruments. His message was clear: AI’s brilliance shouldn’t be sufficient; it should even be reliable.
Persevering with the momentum on DHS 2025, Day 3, Chi Wang, Co-Creator and Co-Founding father of AG2, took the stage with MassGen: Scaling AI By Multi-Agent Collaboration. His stay demo – a number of AI brokers fixing issues collectively – was a vivid showcase of how collaboration, not scale alone, will outline the following frontier of AI. For a lot of within the viewers, this was the primary glimpse of how multi-agent techniques can mirror human teamwork, converging on superior options by way of collective reasoning.
A Sovereign AI Imaginative and prescient for India
Some of the anticipated voices of the day, Pratyush Kumar, Co-Founding father of Sarvam AI, delivered his keynote, Constructing India’s AI Ecosystem: From Imaginative and prescient to Sovereignty. His name to motion was profound: India should construct not in isolation, however with strategic autonomy. This is able to be sure that computing, knowledge, and innovation stay rooted inside the nation. With a passionate attraction for expertise improvement and home investments, he painted a future the place India shouldn’t be a participant however a shaper of the worldwide AI narrative.
Eliminating Complexity, Elevating Innovation
No session captured the spirit of accessibility higher than Mayank Aggarwal, Co-Founder & CEO of evolvue AI, who led Automate The whole lot: Constructing Agentic AI Workflows with No-Code Instruments. With humor, sensible demos, and real-world case research, he confirmed how no-code isn’t “no energy.” As a substitute, it’s a revolution that places AI improvement within the palms of everybody, breaking limitations and democratizing innovation at scale.
Voice: The New Command
Subsequent got here a session that felt like peering into the way forward for human-computer interplay. Microsoft’s Ranjani Mani, Director and Nation Head – Generative AI, and Manoranjan Rajguru, AI Architect, lit up the stage with AI Voice Brokers: The Way forward for Human-Laptop Interplay. From customer support to healthcare, they demonstrated how voice-first interfaces will redefine communication with machines. From clicks to conversations, how know-how may be made extra human than ever earlier than.
Constructing Belief By Safety and Governance
The dialog then shifted to duty and resilience. Satnam Singh, Chief Knowledge Scientist, and Shivaraj Mulimani, Safety Knowledge Scientist at Acalvio Applied sciences, introduced their groundbreaking session Purple Teaming GenAI: Securing Programs from the Inside Out. They confirmed how anticipating threats earlier than they strike is the one approach ahead. The session unpacked methodologies that probe GenAI techniques for vulnerabilities and fortify them towards adversaries. Their message resonated deeply: within the race to innovate, safety can’t be an afterthought.
Smarter Cities, Smarter Programs
From safety, the highlight moved to infrastructure. Manpreet Singh, Senior Knowledge Scientist at Microsoft, electrified the viewers with Constructing Actual-Time Multi-Agent AI for Public Journey Programs. His demo of RAHAT, an AI system helping railway and airport passengers, gave a glimpse right into a future the place AI-powered public journey techniques are seamless, clever, and accessible. It was innovation designed not for the elite, however for the on a regular basis commuter.
The Battle of the Frameworks
Then got here the second many had been ready for: AutoGen vs CrewAI vs LangGraph: Battle of the Agent Frameworks. Mayank Aggarwal, Praneeth Paikray, and Sanathraj Narayan ignited an electrifying showdown, pitting three agent frameworks towards one another. Every framework showcased distinctive strengths, and the spirited dialogue left attendees buzzing with new concepts. It was a debate that pushed the group to rethink tips on how to construct, optimize, and scale agentic AI.
When Brokers Converse the Identical Language
Avinash Pathak, Senior AI Engineer at NVIDIA, carried the baton ahead with Agent to Agent Protocol: Advantages and Workflows. His discuss was a imaginative and prescient of the long run the place AI brokers can seamlessly talk and collaborate throughout techniques. By showcasing how interoperability unlocks effectivity, he left the viewers with a strong takeaway: the long run is about smarter ecosystems.
Trade on the Intersection of AI and Influence
The business use instances on Day 3 have been simply as inspiring because the frameworks. Prakalp Somawanshi, Principal AI Lead at Shell, in his session Multi-Modal GenAI for Vitality Infrastructure Inspection Reviews, showcased how combining imaginative and prescient and language fashions can remodel inspections within the vitality sector, enhancing security, effectivity, and compliance.
Equally compelling was Pradeep Kumar, Senior Software program Engineer at Emirates, who introduced From Imaginative and prescient to Motion: Multi-Modal Agentic AI in Actual-World Use. His session proved that this isn’t about futuristic goals, however about AI techniques already making tangible affect, from aviation to logistics to retail.
Creativity, Collaboration, and Past
The ultimate periods carried the viewers into the guts of creativity and collaboration. Dhruv Nair, Machine Studying Engineer at Adobe, explored Inventive AI Brokers: Open Instruments for Collaborative Media Creation. He confirmed how AI can turn into a co-pilot for designers, storytellers, and creators. In the meantime, Luis Serrano, Founder and Chief Training Officer at Serrano Academy, captivated attendees with A Visible Information to Consideration Mechanism in LLMs. The session turned a posh idea into a chic, visible narrative.
Collectively, these periods reminded the group that AI isn’t just about effectivity. It’s about enhancing creativeness and human potential.
The AV Luminary Awards: Honoring Day 3 Trailblazers
The day closed with celebration and applause because the AV Luminary Awards acknowledged the High 7 AI Group Contributors. These leaders have superior the AI group with their dedication, collaboration, and affect.
Joshua Starmer PhD
Bhavishya pandit
Dipanjan S.
Lakshmi devi prakash
Abhishek divekar
Sarvagya agrawal
Avinash P.
Their recognition was a tribute to the spirit of collective progress that DataHack Summit embodies.
DHS 2025: Closing Reflections
Day 3 of DHS 2025 was a becoming finale to 3 days of imaginative and prescient, vigor, and velocity. From sovereign AI for India to the battle of frameworks, from securing techniques to democratizing improvement, the conversations sparked at DataHack Summit 2025 will echo lengthy after the lights dim.
As a result of DHS isn’t nearly what AI can do in the present day. It’s about what we, as a group, will make doable tomorrow.
Login to proceed studying and luxuriate in expert-curated content material.
