Day 1 of DataHack Summit 2025 set the foundation for India’s most futuristic event. In case you missed it, you can read all about the DHS Day 1 happenings here. On August 21, Day 2 of DHS 2025 built upon it with bold ideas, cutting-edge demonstrations, and a renewed vision for artificial intelligence. With sessions spanning trust, inclusivity, creativity, and robotics, the second day of the summit reminded everyone why DHS is the beating heart of India’s AI revolution.
Table of contents
DHS 2025, Day 2: AI Experts Assemble
The day opened with a keynote speech by Joshua Starmer, Founder & CEO of StatQuest. Just as always, Starmer captured everyone’s attention with his ukulele performance and riveting session on Quantifying Our Confidence in Neural Networks and AI. By showing how confidence can be measured, from simple neural networks to today’s complex LLMs, he provided a framework for trust in systems that too often produce outputs without accountability. His approach reminded the audience that AI’s future is more about reliability than just power.
Carrying that momentum forward, Manish Gupta, Senior Director at Google DeepMind, explored the vast potential and pressing challenges of Inclusive AI. His talk spanned the promise of foundation models like Gemini to their real-world application in the Global South, from personalized learning to agricultural transformations using satellite imagery. His message was clear: AI must not only be cutting-edge, it must be inclusive, fair, and built for billions.
Beyond Data to Discovery
The spotlight then turned to Abhishek Divekar, Senior Applied Scientist at Amazon. He addressed the audience with The Promise and Pitfalls of Synthetic Data Generation. While synthetic data opens doors to privacy-conscious, diverse training sets, he cautioned against careless scale that amplifies hallucinations and biases. His insights grounded the audience in the delicate balance between innovation and responsibility.
Yet another powerful session came from Alessandro Romano, Senior Data Scientist at Kuehne+Zagel. He bridged the gap from LLMs to Agentic AI, showing how multi-agent systems can solve dynamic, real-world problems. Using frameworks like CrewAI, he demonstrated how collaborative, goal-driven agents mark a step-change from static prompt engineering to intelligent orchestration.
The role of the data scientist is evolving rapidly, and hence, a change is crucial. Anand S from Straive shed light on the same, delivering a thought-provoking Hack Session titled “RIP, Data Scientists.” This wasn’t a farewell but a wake-up call, a bold look at how automation and AI are transforming traditional data roles. Automating repetitive data wrangling and creative, strategic problem-solving are the new norm. His message was clear: data scientists who embrace this shift will thrive. For anyone navigating the tides of the data-driven world, this was a blueprint for the future.
Vibe Coding Showdown
The excitement peaked in the Vibe Coding Showdown. Anuvrat Parashar, Anand S, and Ravi RS Nadimpalli went head-to-head in building applications with AI assistants. In their electrifying Hack Panel on “Vibe Coding Showdown: Building Applications with AI Assistants,” the trio displayed collaboration at its finest – a masterclass in blending human creativity with AI precision. For developers and tech enthusiasts in the audience, it offered a glimpse into the future of coding: a dynamic partnership that transforms how applications are imagined and built.
Building Smarter, Faster, Safer
Practical applications with Pankaj Agarwal, Senior Software Engineer (Machine Learning) at Uber, dominated the afternoon sessions. He showcased how RAG-powered search query optimization can transform vague, noisy food delivery queries into precise, intent-aware results. Meanwhile, Dr. Kiran R, VP of Engineering at Oracle, provided a deep dive into Evaluating GenAI Models. Dr Kiran framed the importance of robust metrics, ground-truth datasets, and human-in-the-loop practices in enterprise and healthcare contexts.
Adding to the momentum, Nitin Agarwal, Principal Data Scientist at Toast, and Rutvik Acharya, Principal Data Scientist at Atlassian, explained how the Model Context Protocol (MCP) is streamlining tool integration for agents. Their live demo showcased how MCP can reduce overhead, standardize workflows, and unlock more scalable automation.
Other sessions pushed boundaries further. Anand S, LLM Psychologist with Straive, sparked debate with his provocatively titled talk, RIP, Data Scientists, showing how LLMs can now perform much of the traditional data scientist’s workflow, and what roles remain essential for humans. Krishna Kumar Tiwari, Co-Founder & CTO at Whitler AI, meanwhile, demonstrated how GenAI agents are revolutionizing performance marketing at scale, crafting millions of personalized creatives in real time.
Pushing the Creative Edge
From Web3 to sustainability, Day 2 proved AI’s versatility. Data scientist Rohan Rao showed how GenAI is reimagining the Web3 ecosystem, powering NFTs, smart contracts, and crypto security. Ruchi Awasthi, Machine Learning Engineer – CTO Office at Pure Storage, tackled the pressing question of sustainability in her talk on LLM Compression. Ruchi broke down techniques like pruning, quantization, and knowledge distillation to make AI more efficient and eco-friendly.
Hitesh Nayak, Senior Director of Data Sciences at Decision Foundry, explored how MCP-driven agents can transform marketing. Harshad Khadilkar, Lead Data Scientist at Tata Group, took a creative turn with his session ‘LLMs Are Boring. How Can We Make Them More Interesting?’ His call for directed randomness and novelty reminded the audience that discovery, not just summarization, should be the north star of GenAI.
The narrative deepened with Pavak Biswal and Abhilash Kulkarni of Dentsu Global Services, who unveiled Saving Ananya, a brand AI playbook for customer retention. Through a demo-driven story of predictive capabilities, risk scoring, and agentic triage, they showed how brands can transform pain points into loyalty. Complementing this, Karan Sindwani, Senior Applied Scientist at AWS, revealed how GenAI is powering production-ready solutions across industries. From customer service to cricket analytics, GenAI is proving to be the art of possibilities.
From Robotics to Full-Stack Agents
One of the day’s most inspiring moments came from Logesh Kumar Umapathi of BLACKBOX.AI, who presented From Language to Robotics. His live demo with robotic arms illustrated how LLMs and reinforcement learning are converging to make AI truly embodied. The audience witnessed firsthand how theory translates into tangible, physical impact.
Equally forward-looking were talks by Daksh Varshneya, Senior Product Manager at Rasa. Daksh introduced Process Calling as a superior paradigm for conversational agents. Saurav Agarwal, Solutions Architecture and Engineering at NVIDIA, showcased the Full-Stack Agentic AI toolkit to accelerate, evaluate, and scale agent systems. Miguel Otero Pedrido, ML Engineer and Founder at The Neural Maze, then took the audience beyond PoCs with his speech. He demonstrated how to productionize agentic systems with robust monitoring and LLMOps practices.
Closing out the technical track, Ayush Thakur, Machine Learning Engineer at Weights & Biases, unveiled The Missing Piece of AI Apps: Evaluation. His hands-on strategies for building evaluators, aligning them with human preferences, and scaling bias analysis proved that evaluation is a feature and not an afterthought.
DHS 2025: Day 2 Award Winners
Amid industry leaders, the evening was also marked by celebration. AV Luminary Awards honoured the Top 7 GenAI Scientists for their groundbreaking contributions:
- Abhishek Divekar
- Amit Sharma
- Karan Sindwani
- Logesh Kumar Umapathi
- Mayank Baranwal
- Sunny Manchanda
- Tanika Gupta
Each recipient embodied the innovation and dedication driving India’s AI movement forward. The hall filled with applause for the collective brilliance and spirit of this community.
DHS 2025: A Summit of Ideas, Energy, and Impact
Day 2 of DataHack Summit, or DHS 2025, was a peek at how the AI community is shaping the future. From confidence estimation in neural networks to embodied robotics, from sustainability in LLMs to agent orchestration at scale, the sessions illustrated a remarkable spectrum of progress.
With two days still to go, the summit’s journey has only just begun. Stay tuned to this space as India’s most futuristic event further unfolds.
