Next-gen CI/CD infrastructure with AI-driven optimization and performance SLAs. Developing a parallel CI architecture (KADC), AI-native Rust CLI (Dodo) for workflow generation, and predictive runtime SLAs. Achieved up to 5.93x build speed improvements with caching.
Universal memory infrastructure to make AI products stateful. Implements chunk-based memory graphs, dynamic context injection, and interoperable sync across AI products with pluggable local/cloud storage.
Intent-native prompt system for seamless cross-device actions without app context-switching. Privacy-first with E2EE OTP exchange, Go + Redis backend, and Flutter/React UIs. Handles 100 concurrent users with <750ms P95 latency.
AI-native CLI integrating multi-agent systems for intelligent code editing. WIP: Olm memory control system for layered memory, garbage collection, and inter-agent conflict resolution. Achieved sub-150μs updates in benchmarks.
Custom version control system in Go with Merkle-tree content addressing, compression, modular file indexing, and dynamic branch management. 2,000+ LOC implementation with documentation.
Real-time speech-to-translated-speech system with Flutter UI, Rust audio engine, and Gemma 3n NLP. On-device pipeline preserves tone, emotion, and code-switching.
Full-stack Android app for parsing financial documents. Built with Go backend, Python OCR, Flutter frontend, and Gemini API integration.
AES-128 CBC implementation with custom padding, cryptanalysis of classical ciphers, and exploration of side-channel attack simulations.
cGAN-based face frontalization model using a U-Net generator and PatchGAN discriminator, with pose-aware preprocessing and hybrid loss functions.