Skip to content
PIndependent Projects

Thingsthat run.

Self-directed work, built outside any employer. Each one exists because a specific problem needed a working answer — a data structure, a database engine, or an ingestion path that holds its latency budget.
Projects
3
Employer
None — independent
P.01Systems

High-performance dataframe library

01IndependentSystems ProgrammingJan 2024 – Nov 2024

High-Performance DataFrame Library

Written in pure C.

A memory-efficient columnar data processing library written in pure C with zero external dependencies, built for embedded systems deployment — storage, indexing and allocation designed together rather than inherited from a runtime.

  • Zero external dependencies, targeting embedded systems deployment.
  • O(n log n) sorting and O(1) column access via hash-indexed maps.
  • Benchmarked on 500K-row datasets with sub-linear memory growth.

Capabilities

  • Columnar storageValues of one type held contiguously
  • Hash-indexed mapsO(1) column access
  • O(n log n) sortingOptimised comparison sort
  • Custom memory allocatorAllocation shaped to access pattern
500K
row datasets benchmarked
10×
faster I/O than Pandas baseline
95%
code coverage
40%
reduction in fragmentation
P.02Data Systems
02IndependentDistributed SystemsAug 2024 – Oct 2024

Scalable Sensor Data Pipeline & Analytics Platform

PythonFlaskPostgreSQLRedisDocker

A distributed microservices architecture for real-time IoT sensor ingestion, built so the latency budget holds as volume grows — with an API gateway and anomaly detection layered over the stream.

  • RESTful API gateway with authentication, rate limiting and caching.
  • Containerised with Docker Compose for horizontal scalability and CI/CD integration.
  • Time-series anomaly detection using statistical methods and sliding-window algorithms.
10,000+
events / minute
<100ms
p99 latency
60%
reduction in false-positive alerts
P.03Research Systems
SYSTEMSAug 20251 stars

NeuroDBAI-Native Hybrid SQL + Vector Database in C

From-scratch database engine in C that unifies relational queries and vector search. Features learned indexes and an RL-inspired query optimizer with SIMD intrinsics.

  • Unified SQL + ANN query execution
  • HNSW-like graph search, B+Tree indexes
  • Zero external dependencies
CDatabaseVector SearchHNSWSIMD
SQL+ANN
unified queries
SIMD
CPU-optimized
0-deps
minimal dependencies