AeroSynth 3D / Single-Pass Drone Video Reconstruction
Turn an aerial drone video into a metric 3D reconstruction through a single continuous flight pass.
Local AI, software systems, and API experimentation. Each project describes its implementation and the evidence available for it.
Turn an aerial drone video into a metric 3D reconstruction through a single continuous flight pass.

Bring local AI chat, model management, documents, coding tasks, and an integrated browser into one desktop application.
I built a React and Electron desktop app around a Python/FastAPI backend, with seven curated GGUF profiles and separate services for chat, documents, memory, tools, and controlled coding tasks.
React/TypeScript UI -> Electron IPC and FastAPI -> chat/tool services -> native GGUF runtime. Qwen3.5 and reasoning profiles use llama-server; other profiles can use llama-cpp-python. Electron webviews power Meera Browser, while a separate Playwright service handles backend browser automation.


Local source reviewed on 15 September 2026. All 22 routing/workspace/task/process/scanner/knowledge regression tests passed. Logs record Qwen3.5 local generation, and generated document artifacts are present. Browser and vision behavior has not been validated end to end in this review.
Seven profiles are configured, not seven independently evaluated model releases. Model Studio uses Pro/Max aliases for Meera/Vision and lists different Max quantizations from the backend policy. The live BF16 smoke run failed; memory cleanup and one role test also fail. No universal VRAM, speed, privacy, autonomous-task success, or signed macOS release guarantee is established.
Develop an ML-based market-analysis and prediction application around ticker-driven historical data and technical indicators.
Create an AI/Robotics learning platform that makes technical learning more interactive.
Use computer vision for real-time industrial monitoring and analysis, including the Jyoti CNC problem context explored during an Intel AI program.

Automate YouTube Music tasks through a custom desktop interface.
Automate home/office tasks using connected embedded hardware.
Experiment with an API-oriented bridge around AI/web capabilities.
Personal software project being actively developed around application routing/integration.

Translate a CLI order request into a validated, signed testnet request while preserving uncertainty about remote outcomes.
Make Vidit’s engineering background, project evidence, navigation, social links, and contact paths accessible through a polished public website.