Systems, trading and ML projects, built from scratch and measured
Systems, trading and ML projects, built from scratch and measured
Farhan Sadeek
HOME
TAGS
ARCHIVES
ABOUT
WRITING
CODE
SEARCH
search
Browse by topic
Tags
27 tags across every write-up.
autograd
backtesting
c
causal-inference
cpp
crypto
databases
distributed-systems
experimentation
grpo
interpretability
llm
market-microstructure
numpy
operating-systems
python
pytorch
quant
reinforcement-learning
research
risc-v
simulation
statistics
storage
testing
trading
transformers
autograd
smolgrad, a NumPy Autograd Engine Checked Against PyTorch
backtesting
A Backtester That Refuses to Run on Leaky Features
c
A RISC-V Kernel That Tests Itself on Every Boot
causal-inference
Why Peeking Breaks A/B Tests, and What to Use Instead
cpp
A Crash-Safe B+ Tree That Survives kill -9
Fuzzing a Matching Engine Against a Brute-Force Twin
Group Commit in a Single-Threaded Redis Clone
What Durability Costs a Kafka-Style Log Broker
crypto
Testing Crypto Reversal Under a Pre-Registered Protocol
databases
A Crash-Safe B+ Tree That Survives kill -9
Group Commit in a Single-Threaded Redis Clone
distributed-systems
What Durability Costs a Kafka-Style Log Broker
experimentation
Why Peeking Breaks A/B Tests, and What to Use Instead
grpo
GRPO on Qwen2.5-0.5B and the Prompts That Teach Nothing
interpretability
Finding GPT-2's Induction Heads in One Forward Pass
Testing GPT-2's Induction Heads and the Backups Behind Them
llm
GRPO on Qwen2.5-0.5B and the Prompts That Teach Nothing
market-microstructure
Fuzzing a Matching Engine Against a Brute-Force Twin
Measuring What Informed Traders Cost a Market Maker
numpy
smolgrad, a NumPy Autograd Engine Checked Against PyTorch
operating-systems
A RISC-V Kernel That Tests Itself on Every Boot
python
Measuring What Informed Traders Cost a Market Maker
Why Peeking Breaks A/B Tests, and What to Use Instead
pytorch
Every Line of a Small GPT, Written by Hand
smolgrad, a NumPy Autograd Engine Checked Against PyTorch
quant
A Backtester That Refuses to Run on Leaky Features
Testing Crypto Reversal Under a Pre-Registered Protocol
reinforcement-learning
GRPO on Qwen2.5-0.5B and the Prompts That Teach Nothing
research
A Backtester That Refuses to Run on Leaky Features
Testing Crypto Reversal Under a Pre-Registered Protocol
Finding GPT-2's Induction Heads in One Forward Pass
Testing GPT-2's Induction Heads and the Backups Behind Them
risc-v
A RISC-V Kernel That Tests Itself on Every Boot
simulation
Measuring What Informed Traders Cost a Market Maker
statistics
Why Peeking Breaks A/B Tests, and What to Use Instead
storage
A Crash-Safe B+ Tree That Survives kill -9
Group Commit in a Single-Threaded Redis Clone
What Durability Costs a Kafka-Style Log Broker
testing
Fuzzing a Matching Engine Against a Brute-Force Twin
trading
Fuzzing a Matching Engine Against a Brute-Force Twin
Measuring What Informed Traders Cost a Market Maker
transformers
Every Line of a Small GPT, Written by Hand
Finding GPT-2's Induction Heads in One Forward Pass
Testing GPT-2's Induction Heads and the Backups Behind Them
