My first ever tech meetup, absolutely worth it, there is no senior junior level, everyone shared their experience, it broke my ice in talking with others. Experienced people shared their journey, it was like wholesome talk, where I didn't feel like meeting everyone for the first time. Each one of them made me comfortable and shared their insights.
DEV-ON SOFTWARE - a service based company, organized this event, who is one step ahead of other companies incase of AI integration in their projects. They build an AI model in SDLC and in DevOps, which can understand the project needs and provide solutions, instead of putting the confidential case studies in Open-ai platforms, they build their own models to assist.
One of the interesting part of the event was about the presenter Subal, who did presentation for the first time on "Architecting for Scale - from hundreds to Millions : Scaling a backend system, one bottleneck at a time". He just came started presentation with a funny reel but with strong message which states that "A viral gym meme highlights a hyper-focused lifter losing a competition to a relaxed, casual gym-goer who skips days but stays stress-free. The reel humorously proves that obsessing over perfect routines and flashy gear leads to quick burnout and failure. Ultimately, it conveys that sustainable, imperfect consistency over the long term will always beat short-term perfection." - which means showing up on something consistently matters more than working on with high intensity.
And more interesting part was, the way he delivered the content, no much deep or technical talk, just a casual talk which covers all the details, in between he forgot few steps, he checked his notes and make it also in a funny way. At this moment, people did not focus much on his way of presentation but on interaction he had with everyone. He treated everyone there was his friends, he was telling whatever the mistake he made, what need to be done and what not to be, everything in detail but not perfect.
This made me understand that, showing up on anything will help us win one day. There were 4 presentations but this stood out, because I could easily relate with me, which gave me confidence of showing up.
Other 3 presentation was completely technical and perfect presentation, where seniors asked more doubts and also shared their work experience on it.
Presentation - 1
Six Architecture Decisions the Open Source Decisions
Dr. Sachin Garg, who has 30 years experience in IT industry, shared his insights on the current generation who builds application with the help of AI without understanding the backend working and meaning of the concept used in the projects. He also added that, understanding architecture is as important as building something.
Also had a discussing about the AI in IT industry, AI Engineer is not just prompting, we have to do some designing architecture, model, LLMS, training data, where to host the entire application, about cloud, Kubernetes, decluster and so on. Prompting is just a part in AI.
Presentation - 2
LLM as Judge & Agent as Judge
This presentation was more technical for me to understand but it was interesting to know more new tech terms.
LLM-as-a-judge is an evaluation method where a separate Large Language Model acts as an automated evaluator to grade the text or responses generated by your primary AI model or application. Instead of relying on manual human reviews or simple statistical word-overlap metrics (like BLEU or ROUGE), you provide the judge model with an input (the user prompt), the candidate output, and a scoring rubric. The judge then outputs a score and a written rationale.
Agent-as-a-judge is an advanced framework that builds upon LLM-as-a-judge. Rather than only assessing a final output, an agent-as-a-judge uses autonomous AI agents equipped with tools to monitor another agent's entire multi-step reasoning process. Because agents can actively run code, verify facts against external environments, and query system states, they are highly effective at evaluating complex, multi-step tasks (like software debugging) where the "journey" matters as much as the destination.
Presentation - 3
Handling Large tool outputs in Agentic Workflows - Oodle AI
The presenter taught about how to do cost cutting and implement new methods to make the process simpler, He was also having 20+ years of experience. He mainly focused on, when there is huge traffic hitting on a site, how to handle it effectively and maintain the billing in an appropriate way to make the process cost effective.
Overall the event taught me, to make a clear structured view of each topic you learn and have a good idea about it. Made few contacts with greater experience and also with different domain people. Also I understood that, both senior and junior have an in-depth knowledge and application oriented perspective on the work what they are doing. I could also see more college students who have strong knowledge on Cloud and AI. Inspiring session! And the overall moto was :
DO ONE THING & DO IT WELL!
Tools suggested : coder.com , kilo.ai
Book Suggested : The Cucukoo's Egg
Top comments (0)