Ramesh Vijayaraghavan 🇮🇳

Learn & Share.
Views are my own & not set in stone.

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@_svs_ If an AI agent is going to talk to candidate and schedule a call etc. what is the experience the candidate goes through and is it desirable?

Eg. I hate all the robo calls and I let my voicemail handle all of them.

What’s the play would be in recruitment?

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@avrldotdev Build a knowledge Cache reading the docs and parameters. Map it to intents and pre-conditions

But I would prefer a deterministic workflows over non deterministic LLM tool selection

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@avrldotdev Step 1: Let LLM think and decide an intent (Preferably normalized).
Step 2: Maintain a metadata of tools which can match the intent. One intent many tools are possibility.
Step 3: Let LLM decide from the filtered tools what to call.

What if it is unseen set of tools?

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@NotNotTushar More detailed writing here

https://t.co/xQfQ7nlnsm

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@NotNotTushar All I am saying is use the right tools for the right job and know the consequences and tradeoffs of these tools.

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@NotNotTushar

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@NotNotTushar I do them day in and day out. You need to be aware of the class of problem and the tradeoffs.

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@bhatnaturally After the fastlane thingy, My tower shows fastlane but network is bad.

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@deepseek_ai Any plans for bulk 24h inferencing at cheaper cost?

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@aditya12anand @_svs_ @p_naix @deepseek_ai Which inference provide you are using and which model in Deepseek ?

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

Ended Amazon prime subscription 🎉

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@hellonehha Interesting question I keep asking is do you really need agents or an automation or rules engine can do the work

Many times enterprise needs deterministic outcomes and don’t need an LLM to decide on which tool to use

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@SumitM_X Authorization header has the format of <scheme> <credentials>

It supports https://t.co/QI4pMjYWsG Schemes

Bearer is one of them. It denotes any one bearing this token can get access to the resource without demonstrating possession of cryptographic key

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@SidJain_80 @SumitM_X what other values are supported for authorization header other than one starting with Bearer

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@kritikatwtss 8:30 PM IST

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@amuldotexe @shreyas @gokulr My problem with 1 member team is there will be more than 1 way to solve it and taste varies. 1 member team can create an echo chamber with LLM

[Grady Booch @Grady_Booch]

Here is my prediction:

The Magnificent Seven - Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla - will continue to push AI down our throats despite diminishing returns to its users, simply because they’ve invested so much they’ve gone past a point of no return.

That notwithstanding, the inevitable Great Implosion will happen to companies such as Anthropic and OpenAI because open source models that can be run locally - with minimal capex and zero risk of surveillance - are quickly and irreversibly becoming good enough.

The whole ”AGI is just around the corner” is a meaningless and silly distraction.

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@vivoplt sessions, tokens and serverside state helps REST APIs remain stateless. They inject the state by externalizing them to the Stateless REST Service.

[Ramesh Vijayaraghavan 🇮🇳 @rvramesh]

@Ayushi_g25 @refsrc Thank you