Hari Teja Murakonda

Current: Principal, AICoE@Coupa Past: Early Employee@Opex Analytics acquired by LLamasoft/Coupa

[Hari Teja Murakonda @hariteja1987]

https://t.co/GUTjcj4iEo
GenAI applications are evolving into agentic workflows — multiple sub agents cross verifying each other to arrive at an optimal answer. These workflows on top of avg model easily beats the best models (GPT4). Chips with high tokens/sec have an edge here.

[Hari Teja Murakonda @hariteja1987]

Interesting how investors are categorizing Generative AI market (cc:“All In” podcast)
Compute(Nvidia, AMD, Grok etc) -> Foundational Models(OpenAI, Anthropic, Mistral etc) -> Tooling (VectorDB, LangChain etc) -> Applications (https://t.co/cnNjSs3VDg, https://t.co/vZJbzvx3rQ etc)

[Hari Teja Murakonda @hariteja1987]

Facing the death of our child, partner, spouse or our parents is probably the greatest pain we will endure in our life time. Terrible to see these lost fathers, mothers, sons, daughters, husbands, and wives. True god are the humans around us - that we deeply connect with.

[Hari Teja Murakonda @hariteja1987]

Error Analysis(EA): ML Model::Appraisal Cycle: Employee. EA, if incorporated early, can help steer your ML dev in the right direction. Helps you address 2 big & common issues -- (1) data drift (2) data/model quality (signal strength, label inconsistencies, model issues, etc)

[Hari Teja Murakonda @hariteja1987]

Wall St folks bought Quandl’s leading indicators at a huge premium. They invented leading signals (1) to infer M&A activity from publicly available data on private jets flying between cities (2) measured Teslas on road by working with insurance firms underwriting it in 2015

[Hari Teja Murakonda @hariteja1987]

I was listening to a podcast on Nasdaq acquiring Quandl. In the past, used its data for my forecasting projects, and clearly, they were ahead of the time. They were extremely good at inventing interesting datasets. For ex,

[Hari Teja Murakonda @hariteja1987]

It is hard for an AI system to replace jobs that rely on strong judgement and domain knowledge. For ex, can an AI system design complex behavioral features that correlate with unusual procurement patterns (root causing). It can guide you, but cannot replace you.

[Michael Watson @MichaelWatsonAI]

@hariteja1987 I agree. I had to get up to speed on sensor analytics this week, and ChatGPT was perfect

[Hari Teja Murakonda @hariteja1987]

ChatGPT saved me >1hr yesterday. Understood below topics in <10mins instead of reading medium/stackoverflow stuff via Google.
1. Stratified sampling using SQL(spark)
2. Cramers V and Biserial correlations
3. ERSOP plans/benefits
4. Unusual procurement patterns

Wow, surreal!

[Hari Teja Murakonda @hariteja1987]

With all these exploding use cases of AI and it’s economic impact, policy makers and economists may soon start looking at: GPU-hours and GPU-hours per capita as a way to evaluate and compare countries progress in AI

[Hari Teja Murakonda @hariteja1987]

This is so true and underappreciated! It helps clean your training sets by identifying inaccurate labels, and messy features

[Andrej Karpathy @karpathy]

When you sort your dataset descending by loss you are guaranteed to find something unexpected, strange and helpful.

[Andrej Karpathy @karpathy]

Potentially nitpicky but competitive advantage in AI goes not so much to those with data but those with a data engine: iterated data aquisition, re-training, evaluation, deployment, telemetry. And whoever can spin it fastest. Slide from Tesla to ~illustrate but concept is general

[Hari Teja Murakonda @hariteja1987]

I was just looking at Croatia’s population (3.9M) which is roughly the size of my hometown in India and 1/3 of Hyderabad’s population. Wow, respect for the nation and its soccer team for what it accomplished in the soccer world-cup! @Croatia_FC_2022

[clem 🤗 @ClementDelangue]

The "software 2.0" post (from 2017?!) will go down as one of the most influential posts of the decade!

We did "take it seriously and tried to build a cool company" @huggingface. Thanks @karpathy!

Re-read it: https://t.co/X1xFqS8nPr
Interview @lexfridman: https://t.co/bT2NN3OE2V

[Hari Teja Murakonda @hariteja1987]

@Uber conversations are fun. Met US army folks, repatriated Afghan/Iraqis for helping US army,VP of a big firm driving for fun,and today was a special one. He apparently drives 12hrs/day to fund a school he is building in Nigeria, and provide free education in his hometown. Wow!

[Ganesh Ramakrishna @ramakrig]

Global supply chains have collapsed in the post-Covid-era.

I have worked with the best brands and best minds in the supply chain space over the last 20 years

A thread to explain why we are seeing massive issues for more than 2 years now🧵

[Hari Teja Murakonda @hariteja1987]

Had pleasant experience renewing license at DMV this week. Used to be a 4hr agony before Covid, now took less than 20m. All paperwork/checks happened prior via web uploads. Feels like Covid helped improve consumer experiences in so many ways through digitization and automation!

[Garry Tan @garrytan]

On my graduation day from Stanford, my parents told me to never let your new credential be the best thing about you.

Good reminder as people secure funding, get into YC, or whatever may be the next brass ring... celebrate the wins along the way, but this is no destination.

[Hari Teja Murakonda @hariteja1987]

@CooksCollision Horrible experience with you guys today. My car was returned with no fuel nor any warning leaving me stranded on the road. Then I drive for 5miles and get "engine overheated stop vehicle". Don't you guys test the car before delivery ???

[Opex Analytics, a Division of LLamasoft @opexanalytics]

An app on .@opexanalytics site on peer bank stress testing
https://t.co/66lMe1IDmi …
#analytics #DataScience #Algorithms
.@hariteja1987