Saptarshi Sarkar

A passionate Software Developer and an open-source enthusiast.

[Saptarshi Sarkar @SSarkar2007]

New article: Maths for DSA

Covered:
• Binary / Hex / Octal
• Bitwise operators
• Sieve of Eratosthenes
• Modular arithmetic
• Euclidean & Extended Euclid
• Newton–Raphson sqrt
Useful for DSA + CP beginners.
Read:
https://t.co/xQcfpWJWyh
#Algorithms #maths

[Saptarshi Sarkar @SSarkar2007]

New article: Maths for DSA

Covered:
• Binary / Hex / Octal
• Bitwise operators
• Sieve of Eratosthenes
• Modular arithmetic
• Euclidean & Extended Euclid
• Newton–Raphson sqrt
Useful for DSA + CP beginners.
Read: https://t.co/xQcfpWJWyh
#algorithm #maths

[Saptarshi Sarkar @SSarkar2007]

Alternative platforms to read from:
• CoderLegion: https://t.co/bxX44ZpPxX
• https://t.co/l3NV7vrweU (@ThePracticalDev): https://t.co/1DOu8tToen

#DSA #Algorithms #Math #Programming

[Saptarshi Sarkar @SSarkar2007]

New article published: Maths for DSA 📘

Covered:
• Number systems & 2's Complement
• Bitwise operators
• Prime checking & Sieve
• Modular arithmetic
• Euclidean & Extended Euclid
• Newton–Raphson sqrt

Essential math for DSA & CP.

Read:
@hashnode: https://t.co/xQcfpWJWyh

[Saptarshi Sarkar @SSarkar2007]

🚀 Check out my latest article!

Demystifying Time and Space Complexity for Beginners

Read it here: https://t.co/jitFMwyqbS

View my full profile: https://t.co/BqdhW1nGsx

[Saptarshi Sarkar @SSarkar2007]

Dropped nd4j for a security CVE and cut bloat.
Result = Fixed the vulnerability AND gained 2300% performance. 🚀

Full source and SimdMatrix class here:
https://t.co/i4KA9lugM5

#Java #SIMD #BuildInPublic #OpenSource #ai

[Saptarshi Sarkar @SSarkar2007]

Can Java handle DL math without heavy native wrappers? ☕️⚡️

Implementing models from scratch & benchmarked a SIMD Matrix class (built using JDK Incubator Vector API) vs. naive loops.

Benchmark (2048x2048):
🔴 Naive: 23s
🟢 SIMD: 0.9s

23x faster. Pure Java. 🤯

[Saptarshi Sarkar @SSarkar2007]

Dropped nd4j for a security CVE and cut bloat.
Result = Fixed the vulnerability AND gained 2300% performance. 🚀

Full source and SimdMatrix class here:
https://t.co/i4KA9lugM5

#Java #SIMD #BuildInPublic #OpenSource #ai

[Saptarshi Sarkar @SSarkar2007]

Can Java handle DL math without heavy native wrappers? ☕️⚡️

Implementing models from scratch & benchmarked a SIMD Matrix class (built using JDK Incubator Vector API) vs. naive loops.

Benchmark (2048x2048):
🔴 Naive: 23s
🟢 SIMD: 0.9s

23x faster. Pure Java. 🤯

[Saptarshi Sarkar @SSarkar2007]

Dropped nd4j for a security CVE and cut bloat.
Result? Fixed the vulnerability AND gained 2300% performance. 🚀

Full source and SimdMatrix class here:
https://t.co/i4KA9lugM5

#Java #SIMD #BuildInPublic #OpenSource #ai

[Saptarshi Sarkar @SSarkar2007]

Can Java handle DL math without heavy native wrappers? ☕️⚡️

Implementing models from scratch & benchmarked a SIMD Matrix class (built using JDK Incubator Vector API) vs. naive loops.

Benchmark (2048x2048):
🔴 Naive: 23s
🟢 SIMD: 0.9s

23x faster. Pure Java. 🤯

[Saptarshi Sarkar @SSarkar2007]

Dropped nd4j for a security CVE and cut bloat.
Result? Fixed the vulnerability AND gained 2300% performance. 🚀

Full source and SimdMatrix class here:
https://t.co/i4KA9lugM5

#Java #SIMD #BuildInPublic #OpenSource

[Saptarshi Sarkar @SSarkar2007]

Can Java handle DL math without heavy native wrappers? ☕️⚡️

Implementing models from scratch & benchmarked a Java Incubator Vector API (SIMD) vs. naive loops.

Benchmark (via JUnit on matrix size 2048x2048):
🔴 Naive: 23s
🟢 SIMD: 0.9s

23x faster. Pure Java. 🤯

[Saptarshi Sarkar @SSarkar2007]

Dropped nd4j for a security CVE and cut bloat.
Result? Fixed the vulnerability AND gained 2300% performance. 🚀

Full source and SimdMatrix class here: https://t.co/i4KA9lugM5

#Java #SIMD #BuildInPublic #OpenSource

[Saptarshi Sarkar @SSarkar2007]

Dropped nd4j for a security CVE and cut bloat.
Result? Fixed the vulnerability AND gained 2300% performance. 🚀

Full source and SimdMatrix class here: https://t.co/i4KA9lugM5

#Java #SIMD #BuildInPublic #OpenSource

[Saptarshi Sarkar @SSarkar2007]

Can Java handle DL math without heavy native wrappers? ☕️⚡️

Implementing models from scratch & benchmarked a custom Java Vector API (SIMD) vs. naive loops.

Benchmark (2048x2048):
🔴 Naive: 23s
🟢 SIMD: 0.9s
23x faster. Pure Java. 🤯

[Saptarshi Sarkar @SSarkar2007]

Dropped nd4j for a security CVE and cut bloat. Result? Fixed the vulnerability AND gained 2300% performance. 🚀

Full source and SimdMatrix class here: https://t.co/i4KA9lugM5

#Java #SIMD #Coding #DeepLearning

[Saptarshi Sarkar @SSarkar2007]

Java is "too slow" for Deep Learning? I tried to investigate it. ⚡️

I’m implementing DL models from scratch and rewrote my matrix ops using the JDK incubator Vector API (SIMD).

Benchmark (2048x2048):
🔴 Naive Loop: 23s
🟢 Java SIMD: 0.9s
That’s a 23x speedup. Pure Java. 🤯

[Saptarshi Sarkar @SSarkar2007]

@andrealbts @langchain4j @andrealbts Thank you. Looking forward to it.

[Saptarshi Sarkar @SSarkar2007]

Woke up to 25k+ impressions on LinkedIn this week. 🤯

The best part? It’s not just vanity metrics. The GitHub traffic to "dl-java-labs" repository spiked immediately (chart below 👇).

Turns out, there is a HUGE silent demand for Deep Learning in Java.

[Saptarshi Sarkar @SSarkar2007]

The ecosystem is heating up with Project Panama, Vector API, and @langchain4j.

I'm building dl-java-labs to prove we can do deep learning natively.

Let's build the future AI ecosystem in Java together! 🚀

🔗 Repo: https://t.co/i4KA9lugM5

#Java #BuildInPublic #AI #OpenSource

[Saptarshi Sarkar @SSarkar2007]

Why Java for AI? ☕️🤖

Enterprises run on Java. @VMwareTanzu says: "If Java ceased to exist tomorrow, civilization would crawl to a halt."

Companies won't rewrite mission-critical systems in Python. They want to bring AI in their Java stack.

Source: https://t.co/hqmNF3tPpN