Aman Sachan

CRAZZYYYY

Kanpur, India Joined on Apr 29, 2026 amansachan92905@gmail.com

Aman Sachan

May 4

Qwen sky proof: compressed memory made a tiny model behave better — with the receipts

#ai #llm #python #benchmarking

1 min read

Want to connect with Aman Sachan?

Create an account to connect with Aman Sachan. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

Aman Sachan

May 4

KVQuant / BitForge: same model, smarter context, better answer

#ai #benchmarking #python #opensource

1 min read

Aman Sachan

May 4

LLM Foundry on a tiny model: the stack still does the heavy lifting

#ai #llm #python #opensource

1 min read

Aman Sachan

May 4

LLM Foundry: why the stack around the model matters more than the model itself

#ai #machinelearning #python #opensource

2 min read

Aman Sachan

May 4

SignalHop: the acoustic mesh networking stack that talks through sound

#networking #iot #python #opensource

2 min read

Aman Sachan

May 3

KVQuant: real terminal proof for KV-cache compression

#ai #llm #machinelearning #performance

5 min read

Aman Sachan

May 3

LLM Foundry finally stops being a toy and starts acting like a system

#ai #llm #python #opensource

3 min read

Aman Sachan

May 3

LLM Foundry: the boring stack that makes an LLM actually useful

#ai #llm #openai #anthropic

3 min read

Aman Sachan

May 3

I built a modem that talks through sound. Yes, literally sound.

#networking #iot #python #opensource

3 min read

Aman Sachan

May 2

GhostPilot: Build a GPS-Denied Drone Navigation Stack with Visual SLAM + Agentic AI

#robotics #slam #drones #opensource

11 min read

Aman Sachan

May 1

DataPop: Generate Realistic Synthetic Datasets in Python

#data #database #python #testing

2 min read

Aman Sachan

Apr 30

GhostPilot: GPS-Denied Drone Navigation with Visual SLAM and Agentic AI

#robotics #ros2 #drone #ai

1 min read

Aman Sachan

Apr 30

KVQuant: Run 70B LLMs on 8GB RAM with 4-bit KV Cache Quantization

#python #llm #quantization #optimization

1 min read

Aman Sachan

Apr 30

BitForge: Run LLMs on Microcontrollers

#llm #esp32 #iot #python

1 min read

Aman Sachan

Apr 30

KVQuant: Run 70B LLMs on 8GB RAM with 4-bit KV Cache Quantization

#python #llm #quantization #optimization

1 min read

Aman Sachan

Apr 30

BitForge: Run LLMs on Microcontrollers

#llm #esp32 #iot #python

1 min read

Aman Sachan

Apr 30

VoxelNav: Real-time Semantic Voxel Mapping for ROS2 at 100ms

#ros2 #robotics #mapping

1 min read

Aman Sachan

Apr 30

KVQuant: Run 70B LLMs on 8GB RAM with KV Cache Quantization

#python #llm #quantization

1 min read

Aman Sachan

Apr 30

BTCRouter: Real-Time Bitcoin Fee Estimation Without a Full Node

#python #bitcoin #cryptocurrency #opensource

2 min read

Aman Sachan

Apr 30

QueryFS - SQL Query Your Filesystem

#python #cli #tools #opensource

1 min read

Aman Sachan

Apr 30

SoundForge - Clone Any Voice in 10 Seconds, Export to C/WASM/ESP32

#ai #audio #embedded #webassembly

1 min read

Aman Sachan

Apr 30

VoxelNav - Real-time 3D Semantic Mapping for ROS2 Robots

#robotics #ros #cpp #ai

1 min read

Aman Sachan

Apr 30

KVQuant: Run 70B LLMs on 8GB RAM with Real-Time KV Cache Compression

#python #llm #ai #opensource

1 min read

Aman Sachan

Apr 30

I Compressed GPT-2 to Run on an Arduino ($3 Microcontroller) — Here's How

#python #machinelearning #opensource #ai

1 min read

Aman Sachan

Apr 30

I Compressed GPT-2 to Run on an Arduino

#llm #embedded #tinyml #python

1 min read

DEV Community

Aman Sachan

Badges

Writing Debut

GitHub Repositories

Qwen sky proof: compressed memory made a tiny model behave better — with the receipts

Want to connect with Aman Sachan?

KVQuant / BitForge: same model, smarter context, better answer

LLM Foundry on a tiny model: the stack still does the heavy lifting

LLM Foundry: why the stack around the model matters more than the model itself

SignalHop: the acoustic mesh networking stack that talks through sound

KVQuant: real terminal proof for KV-cache compression

LLM Foundry finally stops being a toy and starts acting like a system

LLM Foundry: the boring stack that makes an LLM actually useful

I built a modem that talks through sound. Yes, literally sound.

GhostPilot: Build a GPS-Denied Drone Navigation Stack with Visual SLAM + Agentic AI

DataPop: Generate Realistic Synthetic Datasets in Python

GhostPilot: GPS-Denied Drone Navigation with Visual SLAM and Agentic AI

KVQuant: Run 70B LLMs on 8GB RAM with 4-bit KV Cache Quantization

BitForge: Run LLMs on Microcontrollers

KVQuant: Run 70B LLMs on 8GB RAM with 4-bit KV Cache Quantization

BitForge: Run LLMs on Microcontrollers

VoxelNav: Real-time Semantic Voxel Mapping for ROS2 at 100ms

KVQuant: Run 70B LLMs on 8GB RAM with KV Cache Quantization

BTCRouter: Real-Time Bitcoin Fee Estimation Without a Full Node

QueryFS - SQL Query Your Filesystem

SoundForge - Clone Any Voice in 10 Seconds, Export to C/WASM/ESP32

VoxelNav - Real-time 3D Semantic Mapping for ROS2 Robots

KVQuant: Run 70B LLMs on 8GB RAM with Real-Time KV Cache Compression

I Compressed GPT-2 to Run on an Arduino ($3 Microcontroller) — Here's How

I Compressed GPT-2 to Run on an Arduino