Mohanraaj Govindaswamy S

Facilitator @descienceosclub | Connecting industry + students on open-source AI experiments | Looking for collaborators, mentors & founders who build in public

[Mohanraaj Govindaswamy S @mohan_raaj]

I’m the facilitator of @descienceosclub.

We connect:
• Students who want real open-source AI experience
• Developers & technical leaders who experiment and share findings
• Founders looking for research-minded students to guide and collaborate with.

1/2

[Mohanraaj Govindaswamy S @mohan_raaj]

Thank you to everyone who's joined us over 17 Open Source Fridays 💚

To bring you closer, the OSF app is now on Google Play:
- book in one tap
- reminders before we go live
- every episode & speaker

Free, no account needed 👇
https://t.co/wjowOWD9K0

#OpenSourceFridays #dosclub

[Descience Open source club @descienceosclub]

It’s not about avoiding failure. It’s about how fast you bounce back.

Most people don’t fear failure, they fear the time it takes to recover.

Resilience isn't ignoring the hit and pretending nothing happened. True resilience is looking in the mirror, saying “Yes"

#dosclub

[Mohanraaj Govindaswamy S @mohan_raaj]

Now it's race with data transfer speed vs hardware 🤔

[Descience Open source club @descienceosclub]

#Marvell buying #Lumentum for $115B proves a hard truth: AI compute hit the #copper wall.

Scaling 100k+ accelerator clusters with copper wiring burns too much power.

The next frontier of AI is not faster chips.

It is photonics moving directly onto the processor substrate.

[Descience Open source club @descienceosclub]

#Marvell buying #Lumentum for $115B proves a hard truth: AI compute hit the #copper wall.

Scaling 100k+ accelerator clusters with copper wiring burns too much power.

The next frontier of AI is not faster chips.

It is photonics moving directly onto the processor substrate.

[Descience Open source club @descienceosclub]

Layoffs news-ah paathu bayandhrukingala? Don't panic, this is a corporate game works!

Don't let layoff news scare you out of your career goals! Understand the actual economics and market strategies behind corporate hiring cycles.

#dosclub #TechLayoffs #Microsoft #CampusHiring

[Descience Open source club @descienceosclub]

How to Make Your Resume Bypass HR Rejections..!

Why do resumes get rejected instantly? This Friday, we dive into practical fixes!

What you'll learn:
✅ How HR & ATS scan resumes
✅ Making your core skills stand out

Register: https://t.co/LcA7zwX7L4

#Resume #JobSearch #dosclub

[Descience Open source club @descienceosclub]

10,000 applicants for ₹3.5 LPA mass recruitment.
0 developers who know how to navigate an enterprise open-source repo.

A degree gets your resume opened. Public GitHub commits and peer collaboration get you hired

Stop building generic apps.

What's the biggest gap in college?

[Mohanraaj Govindaswamy S @mohan_raaj]

Nearly 90% of enterprises use AI, but only 37% report any positive impact on EBIT

and the top performers account for just 6%. The widely cited MIT Project NANDA report put it even more starkly: 95% of enterprise generative AI initiatives fail to produce measurable financial ROI.

[Descience Open source club @descienceosclub]

Join OSF Ep. 15 for a slide-free talk show with @HCLTech Delivery Lead Mr. Selvaganapathi Ramachandran on "Niche & High-Growth Technical Domains"

Starts @ 7:00 PM – 8:00 PM IST - Entry FREE

Live Open-Mic: https://t.co/0EAYRM2k87

#TechCareers #dosclub #OpenSourceFridays

[Mohanraaj Govindaswamy S @mohan_raaj]

@ArjunVegeta What is the future for SaaS now..??

[Mohanraaj Govindaswamy S @mohan_raaj]

Wait what?

Indra OCR model is out now.

We can recognise handwritten notes now with this 1Million parameter.

[Vik Paruchuri @VikParuchuri]

Indra OCR 1.1 is out now!

- 1M param main OCR model, 44 pages/s on CPU
- Within 4 points of 3.4B param DeepSeek OCR 2 on OmniParseBench
- 15x faster than Tesseract, and more accurate
- 80+ languages, plus handwriting

[Mohanraaj Govindaswamy S @mohan_raaj]

@laraavci_ @artur_wala Waiting for mine 🤞

[Mohanraaj Govindaswamy S @mohan_raaj]

@DilumSanjaya I was waiting for your next experiment.

[OpenAI Developers @OpenAIDevs]

Looking for a weekend project?

Here’s what the community has been building and how you can replicate it yourself:

https://t.co/J7mgdmgt9C

[Mohanraaj Govindaswamy S @mohan_raaj]

Yeah, I am talking to you 🫵

[Mohanraaj Govindaswamy S @mohan_raaj]

A promise was once made, and that promise will be kept. @elonmusk

[Elon Musk @elonmusk]

Turns out the future does have flying cars

[Mohanraaj Govindaswamy S @mohan_raaj]

@CKPillaiAI @descienceosclub Exactly. Cross-system IDs, freshness SLAs, and permission-aware retrieval are the real prerequisites. A tight end to end pilot on one decision usually teaches more than broad autonomy on fragmented sources.

[Mohanraaj Govindaswamy S @mohan_raaj]

Placing an autonomous AI agent on top of fragmented enterprise data and expecting business transformation is an architectural illusion.

The blocker in production agentic AI isn't model reasoning, it's siloed pipelines, poor context retrieval, and missing execution guardrails.

[Mohanraaj Govindaswamy S @mohan_raaj]

The next wave of agent architecture requires shifting from " #tokenmaxxing " to " #valuemaxxing ".

Top engineering teams are moving away from brute-forcing frontier LLMs. They are routing tasks dynamically across open-weight models (Qwen, Kimi, GLM) and enforcing strict token budgets per workflow.