Mahit saxena (@MahitSaxen8) header
Mahit saxena (@MahitSaxen8) profile picture

Software Developer | Full-Stack Engineering | AI/ML | Cloud & DevOps

Lucknow
Joined August 2025
Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)Photo by Mahit saxena (@MahitSaxen8)
What a knock KL rahul @klrahul
20
What model do you prefer for coding every day Be honest
5
Fine-tuning a model to extend context from 8k to 64k tokens? If you don’t scale down the learning rate on the base embedding table, the initial token representations drift radically from their pretrained geometry. Freeze or scale down embedding learning rates, or your short-context reasoning degrades as long-context expands. #Transformers #MachineLearning #AIEngineering
5
gemini neon is too good in @GoogleAIStudio but please provide it in antigravity
13
what is this behavior @GoogleDeepMind @OfficialLoganK @rodydavis provide atleast one model for student pro
2
64
Protecting the database layer isn't an afterthought—it's the core of application security. 🔒 Key priorities:• Parameterized queries to defeat SQLi • Principle of least privilege for roles & IAM • Real-time audit logs & anomalous query detection Where are most teams still failing? #DatabaseSecurity #AppSec #DevSecOps #CyberSecurity #WebDev #Backend #Coding
3
Grateful for this experience and the opportunity to engage with global challenges through technology, creativity, and problem-solving. Looking forward to learning, building, and creating meaningful impact. #NASASpaceApps #GalacticProblemSolver #Innovation #Technology
1
12
will ai pro student pack will ever get gemini frontier models ? @grok
12
Has anyone tested Hydrogen i am hearing very negative review about it @grok verify
1
7
Here for Nasa Space Hackathon
22
Is ai replacing software Engineer's @grok verify
1
2
19
Why antigravity feeling super slow today?
2
33
I've realized I've become a little too dependent on AI for coding.
2
17
When will student pro get argon in antigravity @antigravity @OfficialLoganK @GoogleDeepMind
1
38
What concept took you the longest to genuinely understand in programming, and what finally made it click for you? @X
1
19
If you could only keep 3 tools/extensions in your developer environment for the rest of your career, what are they?
1
41
Be honest: What’s a technology/framework you use daily that you secretly think is overhyped @X
1
4
Most teams set clip_grad_norm_ to 1.0 without thinking. Clipping doesn't fix unstable training dynamics—it merely masks them. If your gradients hit the clipping threshold on >20% of training steps, your learning rate is too high or your architecture has exploding paths. #PyTorch
1
4
Is it me or Everyone`s antigravity is super slow @antigravity @rodydavis
1
54