╔══════════════════════════════════════════════════════════════╗
║ FILE : SD-2026-ALPHA CLEARANCE : OPEN-SOURCE ║
║ SUBJECT : Shashank Dev ORIGIN : Bihar, India ║
║ ROLE : Founder & CEO COMPANY : D-MechatronicX ║
║ ARC : ₹280 Arduino → World Champion → NeurIPS 2026 ║
╚══════════════════════════════════════════════════════════════╝
subject= {
"name" : "Shashank Dev",
"age" : 20,
"base" : "Bihar, India",
"role" : "Founder & CEO — D-MechatronicX · CTO & Co-Founder — Syntheta AI",
"arc" : "₹280 Arduino (2021) → World Champion (2024) → NeurIPS (2026)",
"active_ops" : ["AEGIS-X", "CCPL/NeurIPS", "SH-WFS AO/ISRO", "iDEX DISC14", "Syntheta AI"],
"loop" : ["sense()", "predict()", "act()"],
"doctrine" : "Don't wait for the right conditions. Build through them.",
}╔────────────────────────────────────────────────────────────────────╗
│ SYNTHETA AI // SYNTHETIC DATA INFRASTRUCTURE FOR PHYSICAL AI │
│ ─────────────────────────────────────────────────────────────── │
│ Co-founded with Jay Aditya & Hardit Singh · CTO │
│ │
│ Problem → Real-world data for robots/drones/factories │
│ is scarce, expensive, and slow to collect │
│ Approach → 10–50x synthetic amplification anchored to real │
│ captures, validated against domain gap │
│ Pipeline → Real2Robust mutation · NeuroDepth-T4 low-light │
│ restoration · FunctaGen neural re-synthesis │
│ Status → YC application · 2 signed LOIs · pre-revenue │
│ Site → synthetaai.com │
╚────────────────────────────────────────────────────────────────────╝
ID | SYSTEM | DOMAIN | STATUS | METRICS |
|---|---|---|---|---|
01 | AEGIS-X | Counter-Drone AI |  | 88.7% SR · 13ms · RPi5 + Arduino |
02 | CCPL | Constrained RL |  | Dual-stream actor-critic · Pure NumPy |
03 | SH-WFS AO Pipeline | Adaptive Optics AI |  | C+OpenMP centroiding · CNN reconstruction · SLODAR |
04 | ROADSOS | Emergency AI |  | 60+ countries · crash detection |
› SYSTEMS_MANIFEST --archived (7 additional builds)
ID | SYSTEM | DOMAIN | STATUS | METRICS |
|---|---|---|---|---|
05 | TARS Robot | Offline AI |  | 100% offline · IMM-EKF · episodic memory |
06 | RoboticOS | Embedded OS | | 11 subsystems · RTOS-inspired · safety kernel |
07 | Nexus MultiAgent | Multi-Agent AI | | 6 agents · 12 tools · ChromaDB memory · Ollama |
08 | Chronos Engine | Causal Simulation | | Fork realities · causal timeline engine |
09 | bird-language-ml | Bio-Acoustics AI | | EfficientNet-B0 · species + call intent · FastAPI |
10 | Gugugaga | Audio ML | | Infant cry classifier · EfficientNet-B0 · mel-spec |
11 | waifu | Dev Companion | | Tamagotchi for devs · persistent memory |
╔────────────────────────────────────────────────────────────────────╗
│ CCPL // Causal-Consequence-Penalized-Learning │
│ ─────────────────────────────────────────────────────────────── │
│ Most RL agents chase reward. │
│ CCPL agents learn to care about what happens next. │
│ │
│ Architecture → Dual-stream actor-critic │
│ Penalties → State-conditioned, not hardcoded │
│ Delay → Latent estimation, built in │
│ Multipliers → Lagrange, learned — not tuned │
│ Backend → Pure NumPy · zero ML frameworks │
│ Venue → NeurIPS 2026 │
│ Note → Joshua Achiam (OpenAI) engaged directly on math │
╚────────────────────────────────────────────────────────────────────╝
╔────────────────────────────────────────────────────────────────────╗
│ AEGIS-X // COUNTER-DRONE DECISION ENGINE │
│ ─────────────────────────────────────────────────────────────── │
│ Not a drone detector. The decision layer above all hardware. │
│ │
│ Intercept Rate → 88.7% SR │
│ Decision Latency → 13ms │
│ Hardware → Raspberry Pi 5 + Arduino │
│ Algorithms → IMM-EKF tracking · APN guidance · Hungarian │
│ Version → v18 │
│ Submission → iDEX DISC14 │
╚────────────────────────────────────────────────────────────────────╝
╔────────────────────────────────────────────────────────────────────╗
│ SH-WFS AO PIPELINE // ADAPTIVE OPTICS FOR SPACE IMAGING │
│ ─────────────────────────────────────────────────────────────── │
│ Shack-Hartmann wavefront sensing + AI-driven correction. │
│ │
│ Centroiding → C extension + OpenMP (parallelized) │
│ Reconstruction → CNN-based wavefront estimator │
│ Turbulence → SLODAR characterization │
│ Geometry → Fried actuator alignment │
│ Submission → ISRO BAH 2026 · Challenge 9 │
╚────────────────────────────────────────────────────────────────────╝
[2026] Co-founded Syntheta AI — synthetic data infra for physical AI
[2026] ISRO BAH 2026 submitted — SH-WFS adaptive optics pipeline
[2026] iDEX DISC14 submitted — AEGIS-X counter-drone AI
[2026] NeurIPS 2026 submitted — CCPL constrained RL framework
[2026] Joshua Achiam (OpenAI) engaged directly on CCPL math
[2026] Founder & CEO — D-MechatronicX
[2025] ROADSOS v3.1.1 LIVE — emergency AI, 60+ countries
[2025] 5+ parallel systems shipped in 12 months
[2025] CCPL formally proven, simulation-validated
[2024] 🏆 WORLD CHAMPION — Technoxian · FPV Drone Rescue · Delhi
[2024] Quarterfinalist — Robo Soccer · World Championship
[2024] Bihar's first team on a world robotics stage
[2023] Smart Safety Protector — India Book of Records · IJNRD
[2021] ₹280 Arduino · the loop started
✓ Research collaborations · safe RL · constrained decision-making
✓ Defense-tech conversations · autonomy · counter-drone systems
✓ Physical AI data infrastructure · Syntheta AI partnerships/pilots
✓ iDEX / DRDO / ISRO / Indian defense + space ecosystem
✓ Brutal feedback on CCPL — if you can break it, tell me
✗ Waiting for the right conditions
