NeuronicJan 2026 - Mar 2026
React, FastAPI, PostgreSQL, Redis, Celery, AWS
- Deployed an AI-native learning platform on AWS with an async FastAPI backend behind an ALB, Redis-backed session and embedding caches, and a React 19 SPA on CloudFront, maintaining a sub-200ms p95 API latency.
- Built a hybrid retrieval engine combining fastembed vector search with full-text PostgreSQL indexes, powering the automatic knowledge graph construction across the full note corpus.
- Designed a multi-tenant collaboration system with RBAC, edit-suggestion workflows, real-time group feeds, integrated with Google Calendar with OAuth 2.0 and a webhook event bus.
whitematterNov 2025 - Feb 2026
C++, Python, FastAPI, Next.js, CUDA, AWS
- Built a deep learning framework in C++ with automatic differentiation, 20 layer types (Conv2D, MultiHeadAttention, LSTM) and GPU backends for Metal and CUDA.
- Implemented SIMD-optimized tensor ops (AVX/NEON) and OpenMP-parallelized GEMM hitting 99%+ MNIST accuracy in 3 epochs.
- Designed a browser-based training platform where users describe a model in natural language, agents generate the model architecture, and the system compiles and runs optimized C++ training code with real-time loss/accuracy streaming via SSE.
- Shipped as a multi-stage Docker container with a FastAPI job queue, distributed training workers, ONNX export, mixed-precision FP16, and one-click deployment to AWS EC2 GPU instances for inference saving.