DEV Community

Sarthak Agrawal profile picture

Sarthak Agrawal

404 bio not found

Location Lucknow, India Joined Joined on  github website

Work

CSE Junior at MIT, Manipal

Trace one tensor from model math to serving cost
sarthakagrawal927 profile

Trace one tensor from model math to serving cost

#machinelearning #performance #learning #ai
Comments
1 min read
AI pair-programming became useful when every idea met a benchmark
sarthakagrawal927 profile

AI pair-programming became useful when every idea met a benchmark

#ai #programming #testing #devlog
Comments
1 min read
Two days of kernel debugging ended at one default value
sarthakagrawal927 profile

Two days of kernel debugging ended at one default value

#machinelearning #debugging #testing #devlog
Comments
1 min read
A public roadmap should show which predictions aged badly
sarthakagrawal927 profile

A public roadmap should show which predictions aged badly

#opensource #devlog #productivity #ai
Comments
1 min read
Write the reference before debugging the shader
sarthakagrawal927 profile

Write the reference before debugging the shader

#webgpu #machinelearning #testing #devlog
Comments
1 min read
The browser-training speedup was a curve, not one headline number
sarthakagrawal927 profile

The browser-training speedup was a curve, not one headline number

#webgpu #performance #machinelearning #devlog
Comments
1 min read
The more aggressive matmul kernel lost to the register budget
sarthakagrawal927 profile

The more aggressive matmul kernel lost to the register budget

#webgpu #performance #machinelearning #devlog
Comments
1 min read
Memory64 moved the browser ceiling and exposed an ABI gap
sarthakagrawal927 profile

Memory64 moved the browser ceiling and exposed an ABI gap

#webassembly #webdev #performance #devlog
Comments 2
1 min read
A learning roadmap is stronger when every lesson touches the system
sarthakagrawal927 profile

A learning roadmap is stronger when every lesson touches the system

#learning #machinelearning #opensource #devlog
Comments
1 min read
The fine-tune looked better because the evaluation was broken
sarthakagrawal927 profile

The fine-tune looked better because the evaluation was broken

#machinelearning #ai #playwright #devlog
Comments
1 min read
A perfect narrow score can still be the wrong general model
sarthakagrawal927 profile

A perfect narrow score can still be the wrong general model

#machinelearning #ai #testing #devlog
Comments
1 min read
DPO fixed the collapse and still failed the actual task
sarthakagrawal927 profile

DPO fixed the collapse and still failed the actual task

#machinelearning #ai #opensource #devlog
1
Comments
1 min read
A retry decision was enough to close the first factory loop
sarthakagrawal927 profile

A retry decision was enough to close the first factory loop

#machinelearning #ai #opensource #devlog
Comments
1 min read
One token system fixed a site that felt like two products
sarthakagrawal927 profile

One token system fixed a site that felt like two products

#webdev #design #opensource #devlog
1
Comments 1
1 min read
PostTrainLLM vs Axolotl: Mac specialist workflow or accelerator framework?
sarthakagrawal927 profile

PostTrainLLM vs Axolotl: Mac specialist workflow or accelerator framework?

#machinelearning #ai #osdc #programming
1
Comments
2 min read
PostTrainLLM vs Unsloth: start with the hardware and the decision
sarthakagrawal927 profile

PostTrainLLM vs Unsloth: start with the hardware and the decision

#machinelearning #ai #opensource #programming
1
Comments
2 min read
CodeVetter vs Greptile: repository-aware review or task evidence?
sarthakagrawal927 profile

CodeVetter vs Greptile: repository-aware review or task evidence?

#ai #testing #programming #devops
1
Comments
2 min read
A small language model specialist should earn one job
sarthakagrawal927 profile

A small language model specialist should earn one job

#machinelearning #ai #tutorial #opensource
1
Comments
2 min read
CodeVetter vs CodeRabbit: review guidance or verification evidence?
sarthakagrawal927 profile

CodeVetter vs CodeRabbit: review guidance or verification evidence?

#ai #testing #programming #devops
Comments
2 min read
An agent trajectory dataset needs more than final answers
sarthakagrawal927 profile

An agent trajectory dataset needs more than final answers

#machinelearning #ai #tutorial #opensource
Comments
2 min read
Meeting notes can stay on the Mac
sarthakagrawal927 profile

Meeting notes can stay on the Mac

#ai #privacy #productivity
1
Comments
2 min read
What CodeVetter's public benchmark proves, and what it does not
sarthakagrawal927 profile

What CodeVetter's public benchmark proves, and what it does not

#ai #testing #programming #machinelearning
2
Comments
2 min read
Evaluate a local LLM for a decision, not a leaderboard
sarthakagrawal927 profile

Evaluate a local LLM for a decision, not a leaderboard

#machinelearning #ai #fastapi #opensource
1
Comments
2 min read
A Mac AI assistant should verify its actions
sarthakagrawal927 profile

A Mac AI assistant should verify its actions

#ai #automation #productivity
2
Comments
2 min read
What belongs in a coding-agent verification evidence bundle?
sarthakagrawal927 profile

What belongs in a coding-agent verification evidence bundle?

#ai #testing #devops #programming
1
Comments 2
2 min read
MLX LoRA fine-tuning: the command is the easy part
sarthakagrawal927 profile

MLX LoRA fine-tuning: the command is the easy part

#machinelearning #ai #tutorial #opensource
1
Comments
2 min read
What makes a Mac voice assistant private?
sarthakagrawal927 profile

What makes a Mac voice assistant private?

#ai #privacy #productivity
Comments
2 min read
AI code review and agent verification are not the same
sarthakagrawal927 profile

AI code review and agent verification are not the same

#ai #testing #programming #devops
2
Comments 3
2 min read
Training your own LLM usually means building a specialist
sarthakagrawal927 profile

Training your own LLM usually means building a specialist

#ai #machinelearning #tutorial #opensource
Comments
2 min read
A screenshot is not enough for a screen-aware Mac assistant
sarthakagrawal927 profile

A screenshot is not enough for a screen-aware Mac assistant

#ai #privacy #productivity
Comments
2 min read
A practical way to verify AI-generated code
sarthakagrawal927 profile

A practical way to verify AI-generated code

#ai #testing #programming #devops
Comments
2 min read
What on-device should mean for a Mac AI assistant
sarthakagrawal927 profile

What on-device should mean for a Mac AI assistant

#ai #privacy #productivity
1
Comments
2 min read
Fine-tuning an LLM on a Mac is easy to start and hard to evaluate
sarthakagrawal927 profile

Fine-tuning an LLM on a Mac is easy to start and hard to evaluate

#ai #machinelearning #opensource #tutorial
Comments 1
3 min read
A coding agent changed the code. What proves the task is done?
sarthakagrawal927 profile

A coding agent changed the code. What proves the task is done?

#ai #testing #opensource #productivity
Comments 1
3 min read
loading...