PREMCHAND
YADAVPREMCHAND YADAV
“Building intelligent systems at the intersection of AI, research, engineering, and real-world impact.”
Data Scientist & AI Researcher · Formal Runtime Verification · Quantum-Tensor Architectures · Bio-Graph Intelligence · High-Stakes Analytics
EXECUTIVE PROFILE & METRICS
Data Scientist and AI Researcher with a background spanning statistical analysis & hypothesis testing, predictive modeling & forecasting, and quantitative decision systems. Experienced across the full analytics lifecycle – EDA, feature engineering, and model building through to validation and deployment – on tabular, time-series, and graph data. Has led quantitative research combining forecasting ensembles, portfolio optimization, and rigorous statistical validation (VaR/CVaR, purged cross-validation, walk-forward backtesting, Monte Carlo stress testing) to surface actionable, decision-ready recommendations. Comfortable operating as an individual contributor and as a leader coordinating cross-functional teams toward a shared analytical goal. Co-author of four peer-reviewed papers (IEEE, arXiv) spanning edge AI, computer vision, and runtime safety verification for autonomous systems. Applied this range across agriculture, finance, and space-systems datasets. Currently pursuing a B.Tech in Computer Science (AI & Machine Learning) with a 9.35/10 CGPA.
Q-TENSORFORMER VS. STANDARD FP16 TRANSFORMER
FRONTIER AI & FORMAL VERIFICATION
Glass Box at Orbit: A Constitutional AI Verification Framework for Trustworthy Autonomous CubeSat Intelligence
Authors: V C Premchand Yadav and collaborators · Published June 2026
Orbital data-center-scale CubeSats require high autonomy with zero human in the loop. This paper formulates a runtime Constitutional AI verification layer that intercepts every candidate action from an onboard policy before execution. We synthesize 6 physics-grounded constitutional constraints and 7 Linear Temporal Logic (LTL) safety invariants to rigorously guarantee no hazardous command reaches spacecraft actuators.
- • C1 (Power Budget): State-of-charge must not drop below 28% reserve.
- • C2 (Thermal Limit): Reaction wheel temp bounded to ≤ 45°C.
- • C3 (Momentum Dump): Wheel speed saturation limited to ≤ 90% rpm.
- • C4 (Pointing Keep-Out): Star-tracker boresight exclusion angle ≥ 35°.
- • C5 (RF Transmission): S-band downlink restricted to valid ground windows.
- • C6 (Compute Throttle): Edge NPU thermal power dissipation capping.
G(battery < 20% → ¬dispatch(payload_compute))G(sun_sensor_saturated → X(¬slew_to_sun))G(thruster_fired → ¬F≤30s(radio_transmit))CONSTITUTIONAL AI LTL MODEL CHECKER (CUBESAT)
□ (attitude_error > 5.0° → ◇[0, 8s] reaction_wheel_desat)Q-TensorFormer: Dynamic Resource-Adaptive Quantum-Tensor Transformer
HuggingFace Open Source Architecture & Model Release
An information-theoretic Transformer that dynamically allocates compute and memory based on continuous token entropy and uncertainty. Couples Nested Tensor-Train factorization with dynamic early-exit routing to slash inference energy by 77.9% across NVIDIA A100, Apple M2 Max, and Intel Xeon, while compressing 4K KV-cache footprint by 93.4% using 4:1 GQA and asymmetric INT4 quantization.
ACE: Deep Learning-Based Anti-Surveillance System
IEEE Conference Proceedings
Presents an edge-native anti-surveillance intrusion and threat detection framework using lightweight Convolutional Neural Networks designed to run completely disconnected from cloud computation in sensitive physical security enclosures.
Federated Learning & Quantum Optimization Frameworks
Frontier AI Research Series
Investigates privacy-preserving distributed learning across heterogeneous wearable nodes alongside QUBO/Ising mathematical formulations of complex combinatorial routing problems.
SYSTEMS ARCHITECTURES
Quantitative Risk & Monte Carlo Valuation Suite
Quasi-diagonal covariance clustering avoiding matrix inversion instabilities under ill-conditioned assets.
GraphCL GATv2Conv & Multimodal ESM-2 Residue Alignment
Osimertinib Quinazoline Derivative
RISKOS
Quantitative Research & Risk Validation Platform
A quantitative research and risk management engine combining foundation time-series forecasting (TimesFM, Prophet, Merton) with regime-conditioned ensembles, portfolio optimization, and rigorous statistical validation.
Q-TensorFormer
Dynamic Resource-Adaptive Quantum-Tensor Transformer
An end-to-end information-theoretic transformer architecture that allocates compute dynamically based on continuous token entropy and uncertainty.
AETHER-RAMI
Molecular Intelligence Platform for Drug Discovery
A computational drug discovery system coupling self-supervised graph contrastive learning, GATv2/GPS Graph Transformers, and multimodal Drug-Protein CLIP embeddings across five disease targets (EGFR, BRAF, CDK2, HIV-1 Protease, AChE).
DEBRIX
Autonomous Orbital Debris Tracking & Swarm Interception
A cyber-physical 'Dock-and-Dump' space debris mitigation platform fusing edge vision with Unscented Kalman Filtering and decentralized swarm intelligence under harsh orbital illumination.
AuditX & AlphaTrack Global
Financial Anomaly Detection & High-Throughput Ledger Ingestion
Dual-engine financial analytics infrastructure: AuditX detects transaction abuse and fraud with Gemini AI and ensemble classifiers; AlphaTrack handles real-time portfolio ingestion across global exchanges.
MindTrace AI
Explainable Cognitive Bias & Logical Fallacy Detector
A research-grade NLP interpretability framework identifying cognitive distortions, hidden reasoning biases, and persuasive manipulation in text with token-level attribution.
MedFusion
Multimodal Clinical Decision Assistant
A clinical AI platform pairing local edge vision (MedGemma) with cloud reasoning (K2-Think-v2) to evaluate medical imaging, calculate clinical risk scores, and generate standards-compliant FHIR R4 exports.
Hybrid Renewable Microgrid Simulation
LSTM Load Forecasting & Levelized Cost Optimization
AI-driven simulation for rural microgrid electrification coupling LSTM seasonal load forecasting with HOMER Pro energy optimization to achieve 24/7 dispatch stability.
Q-RouteX & QADS
Quantum Hypergraph Routing & Hybrid Autonomous Decision Systems
Trust-aware quantum hypergraph routing system for network digital twins and hybrid quantum-classical autonomous decision planners.
TECHNICAL TAXONOMY (12 DOMAINS)
Data Science & Statistics
Exploratory analytics, hypothesis testing, distributions, and predictive modeling
Quantitative Research & Risk Modeling
Financial time-series, regime modeling, portfolio theory, and backtesting
Generative AI & LLM Tooling
Foundation models, RAG vector pipelines, and parameter-efficient tuning
Multi-Agent & Decision Systems
Autonomous multi-agent coordination, search, and reinforcement learning
Deep Learning & Computer Vision
Convolutional, recurrent, and transformer vision architectures with explainability
Graph & Geometric Deep Learning
Relational neural architectures and biological protein language models
Quantum & Quantum-Inspired Computing
Tensor decomposition and variational quantum circuits
Model Optimization & Edge AI
Inference acceleration, quantization, and embedded edge deployment
AI Safety & Governance
Formal runtime safety, policy verification, and constitutional constraints
Programming & Application Development
Full-stack languages, analytical packages, and rapid prototyping
Cloud, MLOps & Software Engineering
Cloud architecture, containerization, and medical coding standards
Simulation & Systems Modeling
Astrodynamics, energy microgrids, and state estimation
TRAJECTORY & CREDENTIALS
Founder
RICE (Revolution In Cultivating Excellence) – AgriTech Initiative
- •Led an end-to-end analytics initiative, transforming complex environmental, soil, and market tabular datasets into predictive recommendation models and actionable business insights for the agricultural sector.
- •Conducted comprehensive Exploratory Data Analysis (EDA) to uncover optimization opportunities; built, validated, and iteratively deployed predictive models using statistical performance metrics.
Product Development Intern
SOUL (Sense Our Ultimate Link)
- •Analyzed user behavior and engagement datasets, applying statistical hypothesis testing to identify root causes and guide AI-driven product decisions.
- •Collaborated directly with cross-functional teams (engineers, designers) in an Agile environment, presenting written analytical recommendations to align statistics with core business and user-experience goals.
Spearheaded collaborative technical workshops focused on rapid prototyping and modern AI engineering tools.
Directed administrative workflows, cross-departmental operations, and student engagement initiatives.
Managed technical and financial allocations for student-led innovation initiatives and venture incubation.
VIT-AP University
Bachelor of Technology in Computer Science (AI & Machine Learning)
ALLEN Career Institute
Higher Secondary Education (Class XI–XII, Science)
Emmaus Swiss High School
Secondary Education (Class X)
Led AI-driven simulations for a hybrid renewable energy microgrid project using LSTM-based load forecasting.
Co-authored Paper 1 of the Project October series on runtime Constitutional AI verification for autonomous orbital CubeSats.
TRANSMISSION TERMINAL
Available for AI research collaborations, quantitative systems architecture, and engineering roles. Transmissions are typically acknowledged within 12 hours.