|
1 | | -### [PocketOrreryXR — 3D Solar System on Android XR](https://github.com/hellosaumil/PocketOrreryXR) |
| 1 | +### [Danger Room Protocol: Phoenix — Advanced Combat Simulation](#) |
2 | 2 |
|
3 | | -##### *Mixed Reality, Virtual Reality, 3D, glTF, Hand Gestures; Kotlin, Android XR, Android Studio; Google Antigravity, LLMs* |
| 3 | +##### *Holographics, Spatial Reasoning, Real-Time Combat Analysis; Ruby-Quartz, Cerebro Core* |
4 | 4 |
|
5 | | -- Designed an `immersive 3D` app for the solar system on `Android XR` using `Jetpack Compose` and `SceneCore` spatial APIs. |
6 | | -- Implemented `spatial gestures` with pinch-to-select or by dragging to move the solar system in `3D` space. |
7 | | -- Developed `planetary mechanics` with `axial` rotation and `orbit` simulation in `Kotlin` via `iterative AI prompting`. |
8 | | -- Integrated the latest `androidx.xr alpha SDK` for `Full Space mode` to enable seamless transitions between `VR/AR` environments. |
9 | | -- `Vibecoded 100%` of the app from scratch using `Google Antigravity`, `Gemini 3 Pro`, and `Claude Sonnet 4.5` LLMs. |
| 5 | +- Designed an `immersive 3D` combat training program for student units using `holographic projection` and spatial APIs. |
| 6 | +- Implemented `spatial gestures` for real-time environment modification, allowing for `dynamic battlefield adaptation` during training. |
| 7 | +- Developed `simulation mechanics` with axial rotation and orbit logic to simulate `Space-level combat` and gravity-defying scenarios. |
| 8 | +- Integrated `Cerebro-link` alpha SDKs to enable seamless transitions between `simulated` and `augmented` reality training modules. |
10 | 9 |
|
11 | 10 | --- |
12 | | -### [Phoneme Recognition and Digit Identification — Deep Speech Recognition](https://drive.google.com/drive/folders/1C59f-HTYvSWZ5iUEIMvxXTJogwpAaULU?usp=sharing) |
13 | | -##### *Deep Learning, Speech Processing, Computer Vision, LSTM, RNNs, Spectrograms; Python, TensorFlow, Keras, AWS EC2* |
14 | | -- Developed an `end-to-end RNN pipeline` for phoneme recognition on `[TIMIT](https://catalog.ldc.upenn.edu/LDC93S1)` and `[TIDIGITS](https://catalog.ldc.upenn.edu/LDC93S10)` using `TensorFlow` and `Keras`. |
15 | | -- Scaled training on `AWS EC2 (t2.xlarge)`, reducing epoch time from `2.75hrs` to `~45s` (`99% speedup`) over local execution. |
16 | | -- Architected a custom `PaddedBatchGenerator` with `Masking layers` to handle variable-length audio without data loss. |
17 | | -- Conducted a `grid-search` over layer widths, dropout, and L2 rates; optimized to a `2-layer LSTM` with `~64% accuracy`. |
18 | | -- Validated performance using `Stratified K-Fold Cross-Validation` to ensure generalization on unseen test data. |
| 11 | +### [Blackbird Avionics — Next-Gen Jet Flight Control](#) |
| 12 | +##### *Avionics, Stealth Tech, Telepathic Integration; Python, C++, ROS* |
| 13 | +- Developed an `end-to-end flight control pipeline` for the `SR-71 Blackbird` reconnaissance jet. |
| 14 | +- Scaled `telepathic interface` processing on `Cerebro EMR`, reducing latency between pilot intent and jet response by `~85%`. |
| 15 | +- Architected a custom `PaddedBatchGenerator` with `Masking layers` to handle variable-length signal data from onboard sensors. |
19 | 16 |
|
20 | 17 | --- |
21 | | -### [GPU-Accelerated Face Recognition & Tracking on Real-Time HD Video](https://drive.google.com/open?id=0B9gQb-9dKj0ubWVWWjFLQWJqRnM) |
22 | | -##### *Machine Learning, Incremental Learning, IPCA, LDA, K-Means Clustering, Haar Features; Python, OpenCV, PyCUDA* |
23 | | -- Engineered `real-time HD` face recognition using incremental `IPCA+LDA` for `continuous updates` to classifier knowledge. |
24 | | -- Boosted to `94.67% accuracy` by implementing `novel frame-skipping` to allocate resources for incremental learning. |
25 | | -- Accelerated training pipeline `speed by 24%` on `NVIDIA GPGPUs` using `PyCUDA/Sci-kit CUDA` optimizations. |
26 | | -- Achieved `34.3% power savings` on low-end `NVIDIA GeForce 525M GPUs` compared to `CPU-only` execution. |
27 | | -- `Open-sourced` the `SEAS-FR-DB` benchmark dataset `(1080p, 30fps)`; `[Published](https://ieeexplore.ieee.org/document/8369529/)` research findings at an `IEEE SysCon 2018 Conference`. |
28 | | - |
29 | | ---- |
30 | | -### [Agricultural Terrains Image Classification on Satellite Imagery](https://github.com/hellosaumil/deepsat-aws-emr-pyspark) |
31 | | -##### *Big Data, ML Pipeline, Image Classification, Satellite Images, PCA, Random Forests; Python, PySpark, AWS EMR, S3* |
32 | | -- Deployed a `distributed ML pipeline` on `AWS EMR (Elastic MapReduce)` using `PySpark` to classify `405k images` from the `[DeepSat Kaggle](https://www.kaggle.com/crawford/deepsat-sat6)` dataset. |
33 | | -- Optimized training time by `~85%`, achieving `92%` accuracy using `PCA` for dimensionality reduction across `6` terrain categories. |
34 | | -- Architected scalable `AWS S3` & `Spark ML pipelines` to process `5.6 GB`, cutting training costs to `<$0.20/run` on `m5.xlarge clusters`. |
35 | | -- Built a `dual-mode framework` for seamless `local-to-cloud` transition, running inference on `81k samples` in `<5 minutes`. |
36 | | - |
37 | | ---- |
38 | | -### [ToDo iOS App - SwiftUI](https://github.com/hellosaumil/ToDo-SwiftUI#) |
39 | | -##### *Mobile App Dev, MVVM, User Interface (UI/UX); iOS, Swift, SwiftUI, Xcode, WidgetKit, AppGroups* |
40 | | -- Designed a `playful minimal iOS app` for `visual progress tracking` using fun shapes, colors, animations, and `user interactions`. |
41 | | -- Secured sensitive tasks with `FaceID Biometric Authentication`, adding `Favorites` and `Search` functionality for easy discovery. |
42 | | -- Orchestrated real-time `state sync` between the main app and Home screen `Widgets` using `App Groups`. |
43 | | -- Migrated legacy `iOS 14` (`SwiftUI 2`) written in 2020 to `iOS 26` (`SwiftUI 5`) using `Google Antigravity` and `Claude Opus 4.5` in 2026. |
44 | | - |
45 | | ---- |
46 | | -### [QpiC: Querying Platform with VM Integration on Cloud](https://drive.google.com/drive/folders/0B9gQb-9dKj0uUF9iVEM4UXBRaEk?usp=sharing) |
47 | | -##### *Cloud Computing, Microservices, VM, Multi-tenancy; Python, Heroku, OWL, Sparql, WebPy, Flask, Xen Server* |
48 | | -- Developed a `WebPy` SaaS platform for `SPARQL` queries on `OWL/RDF` ontologies via `RDFLib`. |
49 | | -- Optimized response times by implementing `LRU caching` with `pickle serialization` to eliminate redundant parsing. |
50 | | -- Engineered `multi-tenant` isolation with auto-provisioning and strict `filesystem-based` access controls. |
51 | | -- Automated `VM load forecasting` using `Linear Regression` to drive resource balancing via `Xen-API`. |
| 18 | +### [Optic-Vison Control System — Precision Energy Discharge](#) |
| 19 | +##### *Optics, Ruby-Quartz Engineering, Materials Science* |
| 20 | +- Engineered a high-precision `energy-control visor` using custom-cut `Ruby-Quartz lenses` for `discrete` power management. |
| 21 | +- Developed a `pressure-sensitive firing mechanism` to enable variable intensity of energy discharges, from `precision slicing` to `wide-field concussions`. |
| 22 | +- Conducted exhaustive `stress tests` on visor frames to ensure durability during `Omega-level` energy output. |
0 commit comments