Vignesh O K @DataNerdCode

software engineer Bengaluru Joined May 2024

Tweets

20
Followers

24
Following

1K
Likes

2K

Rohit Ghumare @ghumare64

4 weeks ago

x.com/i/article/2055…

15 26 296 39K 575

View Details

If you want your OpenClaw or Hermes Agent to be able to have perfect total recall of all 10,000+ markdown files, GBrain is here to help. It's exactly my OpenClaw/Hermes Agent setup. MIT-licensed open source. Hope it helps you build your mini-AGI. github.com/garrytan/gbrain

229 699 6K 1.1M 11K

View Details

Min Choi @minchoi

2 months ago

Less than 48 hours ago, Google dropped Gemma 4. Minds are blown. And people are already coming up with wild use cases. 10 examples:

85 187 3K 982K 4K

View Details

Vignesh O K @DataNerdCode

5 months ago

@kerkez_official @AnfieldAgenda @Corballyred Absolutely horrible crossing ability

0 0 0 7 0

View Details

Melo☔ @Melo_lfc

6 months ago

Gmail @gmail

2 years ago

How many #Gmail accounts do you have? 👀

10K 1K 19K 68.4M 2K

83 3K 58K 1.5M 1K

View Details

Vignesh O K @DataNerdCode

6 months ago

New job! @EY_Spain as an AI engineer

0 0 0 24 0

View Details

Vignesh O K @DataNerdCode

6 months ago

@GunnerPuls Netherlands

0 0 0 4 0

View Details

Liverpool FC @LFC

8 months ago

Liverpool FC fanatic Jack Faint is currently undertaking a remarkable running challenge in India ❤️

92 411 6K 341K 73

View Details

ABC @Ubunta

10 months ago

I used LLMs query on Data Warehouse directly. But the response times were brutal, costs spiraled, and I was constantly worried about models going rogue on production data. New Trial: What if I fed the AI from cache instead of source? - LLMs never touch raw databases. I keep a curated snapshot in DuckDB—think of it as my "AI-ready" version of the data. All the messy ETL, joins, and transformations happen once, offline. Then I serve the good stuff through Redis. The LLM just asks for what it needs from cache. - I make everything addressable. Every piece of data gets a smart key like customer_123_q3_revenue or product_xyz_reviews_july. When the AI wants customer insights, it's not running complex queries—it's just asking Redis for the right key. Sub-100ms responses became my new normal. - I handle cache misses gracefully.. A background job pulls from my DuckDB snapshot, backfills Redis, and the next identical request is instant. The user might wait once, but never twice for the same answer. - Hit rates tell me if my caching strategy is working. Staleness metrics show if my data is fresh enough. Miss costs reveal which queries are expensive. These numbers helped me optimize which data to keep hot vs. cold. - I learned the boundaries. This works brilliantly for analytics, dashboards, and AI copilots where users can tolerate slightly stale data. For real-time transactions or strict compliance scenarios, I still route directly to source systems. Cache-first architecture gave me speed, predictable costs, and peace of mind knowing my models weren't wreaking havoc on production systems. Treat your cache as the interface, not an afterthought.

0 1 5 415 2

View Details

Vignesh O K @DataNerdCode

10 months ago

@PSGINT_ @grok is salah the greatest African player of all time

2 0 0 280 0

View Details

Ryan @RyanLFC0

11 months ago

Summer of 23, you just had to be there 😭😭😭

74 255 6K 482K 238

View Details

ABC @Ubunta

12 months ago

Designing an MCP server for a data-warehouse isn’t trivial—but the payoff is huge. Here’s why I’m convinced this pattern is the future of data access - Natural language = democratized data Non-technical users can now query data using plain English. No more SQL barriers, no more waiting for analysts. Just ask and get answers. - Cache frequently used queries is much easier using MCP tools. You can even create specific MCP functions for expensive queries and use cache more smartly. - Users don't need to juggle different UIs or remember which system has what data. Engineers can route queries based on cost, performance, and governance rules behind the scenes - Deep level Granular governance made simple. Different MCP Server tool sets for different roles: interns see exactly what they should, execs see what they must. Governance is baked in, not bolted on. MCP Server brings the business closer to its data, cuts waste, and hides back-end complexity—letting every team ask better questions, faster.

2 3 20 1K 21

View Details

Vignesh O K @DataNerdCode

a year ago

@Ubunta What would be the solution for this as a Data engineer?

0 0 0 18 0

View Details

Vignesh O K @DataNerdCode

2 years ago

What is Data Modelling? Beginner's Guide to Data Models and Data Modelling youtu.be/CUR6rKrIEGc?si… via @YouTube

0 0 0 46 0

View Details

ABC @Ubunta

2 years ago

Not every form of Data Engineering requires SQL. Throughout my career in Data Engineering, I've found myself writing very little SQL, as the majority of SQL work was handled by Subject Matter Experts. As a Data Engineer, my focus was more on: - Building Data and AI platforms using a wide range of tech stacks tailored to different team needs—each project was unique!Adapted to different technologies as needed - Ensuring Data Security, Served as a primary responsibility, While seemingly straightforward, this aspect proved extremely challenging Required constant vigilance and updates - Developing middleware to enable seamless interaction between various running services. This isn’t traditional Data Engineering, but when I started, the role didn’t have concrete responsibilities. Not every Data Engineer works heavily with SQL; instead, DEs are a subset of Software Engineering, where SQL serves as a general-purpose language to interface with databases.

1 2 21 1K 5

View Details

Vignesh O K @DataNerdCode

2 years ago

Join us at virtual MLOps Community event - Data Engineering for AI/ML - September 12, 2024 6:30 PM IST home.mlops.community/public/events/… #MLOpsCommunity