Anubhav Singh
AI at Weights & Biases, Google Dev Expert in ML & GCP
Kolkata, India
Actions
A seasoned developer since the pre-Bootstrap era, Anubhav has extensive experience as a freelancer and AI startup founder. He authored "Hands-on Python Deep Learning for Web" and "Mobile Deep Learning with TensorFlow Lite, ML Kit, and Flutter." A Google Developer Expert in ML & GCP, he was a co-organizer for TFUG Kolkata and formerly led the team at GDG Cloud Kolkata. Anubhav is often found discussing System Architecture, Machine Learning, and Web technologies.
Badges
Area of Expertise
Topics
AI Insights with Weave by W&B
Learn how W&B Weave is driving deeper understanding of AI apps! In this exciting era of AI agents, Weave provides fantastic tracing, experimentation, and essential guardrails, giving you confidence in your AI applications. Get ready to supercharge your workflow with powerful evaluation tools, incredible debugging capabilities, and comprehensive support for building and managing sophisticated AI agents.
Let's build amazing AI, together!
Welcome Keynote: Upskilling and Building with AI on Google Cloud
In this event welcome keynote, let us take a look at how we can approach upskilling and ideating solutions with AI capabilities and leverage Google Cloud in the journey.
I'll introduce the audience to various Google Cloud AI services and resources while giving an overview of the event to the audience, so that they can make the best of GCCD Kolkata 2024.
Tiny Memory Efficient AI Agents using Golang and Genkit
What if we could make AI agents that were just as powerful but slim enough to run on limited memory, all without draining resources? That’s when I stumbled into the world of tiny, memory-efficient AI agents, and Golang with Genkit became my toolkit of choice.
In this talk, I’ll share the journey of crafting AI agents that make smart use of memory without sacrificing performance. Using Golang’s concurrency superpowers and Genkit’s modular design, I’ve been able to build agents that keep memory usage lean yet powerful enough to handle real-world tasks. This session will walk through how to apply these principles to create AI that’s not only fast but resource-efficient—ideal for edge cases and low-resource environments.
If you’re curious about how minimalism in code can transform AI applications, come join me! I’ll cover hands-on examples and practical insights to show you how we can build intelligent, tiny agents that are surprisingly robust and ready to deploy.
Reimagine what's possible with Generative AI Studio
While a lot of companies are trying to use generative AI for question answering based tasks, with Generative AI Studio on Vertex AI, the possibilties are much more varied. In this session I will be exploring the different ways in which generative AI is available on GCP and how to think AI first to build solutions that go beyond the trend.
The talk will include a comprehensive discussion of all major offerings available in the Generative AI studio on GCP and include my learnings from building a gen ai focused solution at my startup - callchimp.ai.
Goroutines as Cognitive Threads: Replicating Human Behavior in Go
Humans excel at multitasking, and Go's goroutines can replicate this in a humanoid system. In this talk, we'll integrate Gemini with Firebase Genkit Golang, leveraging Google Search for real-time grounding and context caching, to create a highly concurrent system that mimics human cognitive functions. Learn to architect complex task orchestration in Go.
Overview:
This talk will cover:
- The anatomy of goroutines as cognitive threads
- Integrating Gemini using Firebase Genkit Golang
- Using Google Search for real-time grounding and context caching
- Synchronizing tasks and managing shared state
- Case studies and practical examples of humanoid multitasking
Job Scheduling on Google Cloud
This session will explore the different services available in GCP for performing various types of job scheduling. We'll discuss which method to use and when, and show basic demos of each method.
We'll be exploring options for job scheduling via GCP services and third party tools, along with the benefits and pitfalls of the different methods.
Talk Outline:
1. Introduction
2. What is Job Scheduling and its importance
3. Types of Scheduling
3.1. Time-based Scheduling
3.2. Monitoring and Logging
3.3. Event response
3.4. Event-based Scheduling
3.5. Error Handling and Retries
3.6. System Maintenance
4. Methods of Job Scheduling
4.1. Cloud Scheduler
4.2. Cloud Functions
4.3. Cloud Pub/Sub
4.4. Cloud Workflows
5. Demos
5.1. Using cron jobs
5.2. Using framework event triggers
5.3. Using Cloud Scheduler
5.4. Using Pub/Sub
5.5. Using Workflows
6. Conclusion
Cloud Run as a backend for load balancing
In this talk, we'll discuss using Google Cloud Run as a backend for load balancing in the Google Cloud Platform. We'll cover the benefits of Cloud Run, such as automatic scaling, stateless and concurrent processing, integration with Google Cloud services, and cost-effectiveness. Additionally, we'll explore how to set up Cloud Run with Google Cloud Load Balancing for custom domains and SSL termination, enabling you to build highly-scalable and high-performance backend services for your applications.
Cloud is all you need
Cloud Computing has taken the world by a storm - driving businesses and research worldwide, ushering us into an era where the specifications of local devices are rapidly becoming irrelevant.
This talk explores a brief history of Cloud computing and how it has impacted businesses, technologies and daily lives over the last 10 years. We then delve into a case study of a few of the Top 10 Fortune 500 companies to see how they have leveraged cloud for growth.
I will also touch upon the intersection of Cloud, ML and Web and finally show a demo of the same.
Sample PPT here - https://xpri.dev/cloud-is-all-you-need
Scaling Computer Vision: From tutorials to fundraising
Hello world to Computer Vision - most tutorials on the internet don't take you beyond the basic implementations.
In this talk, I will take you through a story of how we built a scalable distributed computer vision platform which helped a startup raise $600,000 of funding. This solution was implemented at National University of Singapore.
Outline:
Introduction to the speaker and topic
Problem statement description
How it started
The roadblocks we hit
How we overcame those roadblocks
Towards fundraising
Q/A
Let’s Design a high availability ML heavy Cloud solution architecture
We love the words Machine Learning and Cloud. They almost go together when we talk about ML research, development and deployment. However, more often than not, your models may perform slow or your system faces downtimes and model degradation a lot. How do you ensure that your ML pipeline doesn’t face downtimes and become a bottleneck in the growth of your business?
At Dynopii, we work with real-time deep learning-based models for our products and services that run 24x7. One of the first challenges we solved was to keep our models running with 0 downtimes despite multiple updates to it every day. I’ll be talking about how to analyze your ML pipeline and how to design an ML Cloud solution architecture that keeps your critical business processes running smoothly through both planned and unplanned outages.
This session will be agnostic to cloud service providers!
Plan More, Code Less - Building Scalable Solutions
This session begins with questioning something we hear a lot today - "write code fast". Yes, we should write code fast, but when?
The session focuses on establishing the idea of solid technical research before making decisions regarding the selection of languages and frameworks in order to achieve scalable and performant solutions.
Student developers today, all over the world are driven by the idea of building solutions using the latest and simplest technologies they can get their hands on. This session advocates that they should instead be technology agnostic while designing their solutions, and instead, choose the right technology for the right purpose.
The session discusses at length how developers can decide the technologies they need for their solutions, keeping in mind that their solution would cater to millions.
The Do's and Dont's of building an AI powered website
Considering the vastness to which web applications can grow, and the strong dependence of nearly every other platform for a backend running as a web-based service, it is important for the backend to be well thought of and properly executed. Artificial Intelligence-based applications even in a Proof of Concept stage are often not blazing fast in responding or take a lot of time to train on the new samples. I shall be discussing tips and tricks to make a backend which does not choke under pressure of a production website.
Headings -
1. AI in the Web - Do you see it?
2. Is your AI model fast enough?
3. The Qualification Checklist of an AI backend
4. How to re-design an AI backend to be fast
5. How to update AI models on the production
AI for Web : Web for AI
Overview:
This talk explores the inter-dependence of the World Wide Web and the field of Artificial Intelligence, tracing both the storylines from their independent routes, to chronologically reaching the current day amalgamation, while discussing the major milestones of their relationship. The talk would be the complete story of the merging together of two fields, the factors that made them mutually beneficial and the projected future.
Outline:
1. Introduction to Speaker and Topic intuition
2. The Roots of Web and AI - Mutual independence
3. The Birth of Semantic Nets - First acknowledgement
4. ELIZA, the ARPANET baby - Gaining popularity
5. Stanford's SUMEX-AIM - Establishment of Partnership
6. The Semantic Web Road map by Tim Berner’s Lee - From the Father’s Pen
7. The Rise of the Crawlers - Web as the Fuel of AI
8. OWL Web Ontology Language by W3C - A standard Pact signed
9. Assistants, Advertisements, Security - Today
10. Looking ahead - The Future
The simple science of unsupervised NLP
Unsupervised machine learning has been an often ignored topic at machine learning sessions due to it's complexity of concepts and algorithms.
In this session, through Natural Language Processing, which is a rapidly emerging trend, also powering the various chatbot AIs making booms in the market, we'll be introducing Unsupervised learning.
Next, we'll have a short demo of building an article recommender system based on the Wikipedia articles a person has read recently.
Cloud Community Days Kolkata 2025 Sessionize Event Upcoming
Google Cloud Community Day Gandhinagar 2025 Sessionize Event Upcoming
Road to AI Community Day Sessionize Event
Gdg gandhinagar devfest 2024 Sessionize Event
Google Cloud Community Day - Pune 2024 Sessionize Event
Google Cloud Community Days 2024 Mumbai Sessionize Event
Google Cloud Community Days Kolkata 2024 Sessionize Event
Google Cloud Community Day Gandhinagar 2024 Sessionize Event
Google I/O Extended 2024 Jalandhar Sessionize Event
DevFest Nashik 23 Sessionize Event
DevFest Chandigarh 2023 Sessionize Event
DevFest Yogyakarta 2023 Sessionize Event
Google Cloud Community Day - Pune 2023 Sessionize Event
GDG Ahmedabad DevFest 2022 Sessionize Event
DevFest Nashik 22 Sessionize Event
Cloud Community Day - Pune Sessionize Event
DSC OMG Sessionize Event
GDG Ahmedabad DevFest 2019 Sessionize Event
DevFest Kolkata 2019 Sessionize Event
DevFest Kolkata '18 Sessionize Event
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top