Avinash Jha | Defence-Tech Innovator & Systems Architect

2025

Flagship Simulation Dashboard Ongoing

NETRA

Python Developer & System Architect

Threat Scoring & Digital Twin System for IED Detection

🛡️ Digital Twin System

Real-Time

Threat Scoring

Live Dashboard

5+

Sensor Signals

Multi-Source Fusion

360°

Situational Awareness

UAV Integration

📍 Route Simulation

🔥 Heatmap Viz

⚠️ Anomaly Detection

📊 Probability Bars

Designed real-time dashboard simulating convoy/patrol routes using Streamlit
Built rule-based anomaly detection logic combining terrain indicators & sensor signals
Created heatmaps, probability bars, and scenario playback for high-risk visualization
Integrated UAV recon concept (APYUAA module) to improve situational awareness
Outcome: Enabled faster decision-making and clearer threat evaluation

Python Streamlit Digital Twin AI/ML

View Simulation View Document GitHub

2026

Flagship AI/ML Ongoing

JATAYU

System Architect & ML Engineer

Predictive Intelligence Fusion System — Threat Pattern Reverse-Engineering & Risk Forecasting Engine

📊 Data Intelligence

200+

Raw IED Incidents

Red Corridor 2020-2026

8.2M

Synthetic Data Points

Pattern Amplification

23

Engineered Features

Temporal + Spatial + Casualty

📋

Raw Intel 200+ incidents

→

🔬

Feature Engineering 23 signals extracted

→

🧬

Synthetic Expansion 8.2M data points

→

🎯

Threat Prediction Hotspot forecasting

🔥 Top Hotspot Districts

1 Bijapur Chhattisgarh 63 incidents 27 killed

2 West Singhbhum Jharkhand 33 incidents 14 killed

3 Narayanpur Chhattisgarh 24 incidents 13 killed

4 Dantewada Chhattisgarh 11 incidents 12 killed

5 Sukma Chhattisgarh 11 incidents 8 killed

Built large-scale predictive intelligence system to anticipate IED threats before they occur
Reverse-engineered hidden attack patterns from 200+ real incidents across India's Red Corridor
Engineered 23+ feature signals: temporal cycles, casualty gradients, spatial density, and seasonal volatility
Combined XGBoost with LSTM to capture both static and time-dependent threat signals
Built interactive Streamlit dashboard for live hotspot exploration and risk probability trends
Outcome: Identified high-risk regions and early indicators — shifting from reactive to predictive threat awareness

Python XGBoost LSTM Streamlit scikit-learn Plotly

View on GitHub View Dataset

2025

🔒 Confidential Ongoing

APYUAA

Concept Designer

Recon UAV – Aerial Support Module

Developed biomimetic UAV concept inspired by avian flight dynamics
Designed FPV-based route-scouting logic with terrain awareness
Integrated recon workflow with NETRA for unified decision support
Outcome: High-level blueprint for multi-layer surveillance

FPV Biomimicry Recon

2025

Ongoing

Route Simulation

Developer

Anomaly Dashboard & Visualization

Built lightweight simulation interface for route data & anomaly indicators
Designed modular UI components: heatmaps and probability bars
Outcome: Improved clarity and rapid prototyping for simulations

Python Streamlit Data Viz

2026

AI/ML Security Ongoing

Bittachon

System Architect & AI Researcher

Rebuilding Digital Trust Beyond Deepfake Detection

🛡️ Trust Verification Engine

💓

Biological Invariance

rPPG Pulse Detection

Invisible heartbeat signature

🔐

Digital Provenance

C2PA Validation

Cryptographic chain of custody

🎬

Physical Consistency

Audio-Visual Sync

SyncNet lip-sync analysis

📊 Trust Timeline — AI Nutrition Label

✓ Verified

⚠ Flagged

? Unknown

90% faster verification Privacy-first Edge AI

Built multi-modal verification engine with three independent pillars: Biological Invariance (rPPG pulse detection), Digital Provenance (C2PA cryptographic validation), and Physical Consistency (audio-visual sync analysis)
Designed "Trust Timeline" — an AI Nutrition Label for media, showing when trust was established or broken
Implemented privacy-first architecture using WebAssembly & Edge AI for local biometric processing
Applied spectral analysis and SyncNet models to detect audio-visual manipulation inconsistencies
Outcome: Paradigm shift from passive deepfake detection to active trust verification — up to 90% reduction in verification time

Python rPPG C2PA WebAssembly Edge AI SyncNet

View Document

2026

Systems Climate-Tech Ongoing

YAROK

Systems Engineer & Industrial Architect

System-Survivable Green Steel Architecture for India

🌱 Green Steel Revolution

75-80%

Electrolyser Utilization

vs typical 30-45%

<0.4

tCO₂e/tonne

CBAM Compliant

₹82-119Cr

Annual O₂ Value

Hidden Revenue Unlocked

⚡ Electrical Buffer

🧪 Chemical Buffer (H₂)

🔥 Thermal Buffer

Designed Hydrogen-DRI steelmaking with Tri-Layer Energy Buffer (electrical, chemical, thermal) achieving 75-80% electrolyser utilization vs. typical 30-45%
Engineered water strategy via Permeate Offtake Model — bypassing ₹200-300 crore desal CAPEX and 18-24 months of approvals
Built ore beneficiation pipeline for Indian hematite (3-5% alumina) using desliming + reverse flotation
Integrated oxygen monetization from electrolysis — unlocking ₹82-119 crore/year hidden value via EAF lancing
Outcome: CBAM-compliant green steel at <0.4 tCO₂e/tonne — turning EU regulation into competitive advantage

H₂-DRI Electrolysis Energy Systems CBAM Process Engineering

View Document

2025

Robotics Ongoing

UZIX

ROS2 Developer & HRI Designer

Natural Language Control of UR3 Robotic Arm

Built hybrid command-parsing architecture — deterministic pattern matching for common instructions + LLM fallback (Gemini) for novel commands
Implemented structured pick-and-place pipeline with explicit motion sequencing and coordinate-frame awareness
Designed gripper-motion synchronization in FalconSim simulation environment
Outcome: Safe, predictable robot behavior with natural language flexibility — minimizing LLM usage while preserving usability

ROS2 Python Gemini API FalconSim NLP

2025

Ongoing

Shadow Twin

System Designer & Graph Algorithm Developer

Live Digital Twin for Tourist-Site Crisis Response & Crowd Management

Building real-time digital twin with dynamic weighted graph model — nodes as zones, edges as paths with crowd-density-updated weights
Implementing event-driven recomputation of evacuation corridors using constrained shortest-path and capacity-aware routing logic
Designing rule-based simulation layer for crowd flow, congestion buildup, and zone saturation modeling
Building human-in-the-loop decision dashboard with structured alerts and decision proposals for operator override
Focus: Low-latency state updates, explainable decision logic, and operational realism for high-footfall environments

Digital Twin Graph Algorithms Real-time Systems Dashboard