Loading portfolio...
I grew up fascinated by two things — mathematics and language. I wanted to understand why equations could predict physical reality and why the same sentence in Hindi and English carried different emotional weight. That curiosity led me to Mathematics & Computing at RGIPT, where I discovered that AI sits exactly at that intersection.
Today, my work clusters around multilingual NLP for code-mixed text, robustness evaluation for vision and VLMs, and applied AI systems such as agents and RAG pipelines. I try to keep the work evidence-backed through datasets, baselines, ablations, error analysis, and reproducible artifacts.
What drives me is the intersection of rigorous mathematics and real-world impact — whether that's reducing LLM hallucinations by 68%, detecting cement plant anomalies from 378K sensor readings, or democratizing AI access for researchers with limited compute.
Currently ranked #1 in Mathematics at my university, with a Minor in Business Consulting that gives me a unique lens on translating technical depth into stakeholder value.
Indian Institute of Management Shillong
Mentor: Dr. Shrawan Kumar Trivedi
Polka Labs Pte. Ltd. (Polkassembly) — Singapore
Sambin
Lizmotors Mobility
AspireNex
Rajiv Gandhi Institute of Petroleum Technology (RGIPT), Amethi
2022 — 2026
Minor: Business Consulting
Key Courses: Machine Learning, Deep Learning, Linear Algebra, Probability & Statistics, Optimization, Data Structures & Algorithms, Operating Systems, DBMS
Leadership: OWASP Management Head — led AI security & blockchain initiatives, guided 500+ freshmen
Delhi Public School, Hapur
2020 — 2022
Delhi Public School, Hapur
2020
Expert Systems with Applications, Volume 308, May 2026
2nd Conference on Applied AI and Scientific ML (CASML 2025)
Reproducible CIFAR-10-C robustness study comparing source-only, BN Adapt, Tent, UG-Tent, and source-mix variants. Source Mix Tent 0.75 reached 82.64% mean accuracy and reduced negative-transfer cases from 5 to 3.
Read study →Zero-shot CLIP ViT-B/32 evaluation on CIFAR-10-C, CIFAR-100-C, and EuroSAT. The study tracks accuracy collapse, calibration error, confidence gap, entropy, and feature drift under corruption and domain shift.
Read study →Adversarial evaluation suite for database-engineering AI agents with golden-database grading, live Mermaid ERDs, and 8 SQLite migration tasks.
View on GitHub →Intelligent AI dispatch system for optimizing task allocation and resource management using multi-agent coordination.
View on GitHub →Real-time anomaly detection system using computer vision and edge computing with YOLOv8 for intelligent monitoring on edge devices.
View on GitHub →Code-mixed Hinglish text-to-speech synthesis system, addressing the challenge of bilingual TTS for India's most common conversational language pattern.
View on GitHub →4 specialized agents (Petrophysical, Seismic, Risk/ESG) with auditable JSON artifacts for 100% reproducibility. RAG-based chat interface that cites sources.
View on GitHub →Forecasting, segmentation & A/B testing on 600K+ pharma transactions across 57 drug categories. Linear/ridge regression, K-means clustering, Tableau/Power BI dashboards.
View on GitHub →Time-series forecasting study on NIFTY 50 data with gradient boosting, leakage-aware feature engineering checks, and 10+ market features including SMAs, RSI, log-returns, and rolling volatility.
View on GitHub →Industrial sensor analytics on 378K+ rows. 28 domain-informed features, Isolation Forest with temporal smoothing, Z-score root-cause analysis. Power BI dashboards.
View on GitHub →Real-time flood prediction with 90% accuracy using geospatial data. Interactive flood maps, evacuation routes, and chatbot serving 40,000+ users.
View on GitHub →85% speaker identification with Siamese networks + SVM. BERT-based embeddings achieving 90% transcription accuracy. Context-aware NLP for speaker turn detection.
View on GitHub →Rebuilt summarization benchmark with ROUGE/compression evaluation, batched HuggingFace inference scripts, Lead baseline, and a NumPy mini Transformer from scratch.
View on GitHub →Boosted conversion rates by 20% and retention by 30%. Collaborative + Content-based Filtering with TensorFlow, handling 1M+ user interactions.
View on GitHub →DQN + Actor-Critic from scratch. 87% win rate CartPole, 65% LunarLander across 50K episodes. Prioritized experience replay, adaptive learning rate scheduling.
View on GitHub →Hybrid Transformer-LSTM + RL for PRNG. Evaluated against NIST statistical test suites on 100M generated bits. Reward function based on entropy and unpredictability.
View on GitHub →Financial analysis tool built for the Algolia MCP Challenge, leveraging AI for automated financial insights and recommendations.
View on GitHub →Interactive visualization of global trade networks, revealing patterns and dependencies in international commerce through data-driven graphics.
View on GitHub →Robustness study on physiological signals using uncertainty quantification and physics constraints. Explores neural network reliability in safety-critical domains.
View on GitHub →Mathematics Department, RGIPT Amethi — CPI: 8.60+
Expert Systems with Applications journal publication + CASML 2025 conference work
IIM Calcutta & IIFT Kolkata — National-level competition, 2024
IIEC IIT Gandhinagar — Built AI platform serving 40,000+ users
Multiple competitive programming platforms in Python & C
Top 1.5 percentile — AIR ~18,247 among 1.2 million candidates
Delhi Public School, Hapur — Outstanding academic record
Led AI security & blockchain initiatives, guided 500+ freshmen at RGIPT
Primary — 65 repositories
4 followers · 37 starsSecondary — 9 repositories
Older projects & web workThis visual is a static portfolio illustration, not live GitHub contribution data.
I actively study and fork repositories from world-class sources to deepen my understanding. These aren't my original work, but they represent my commitment to continuous learning:
I'm always open to discussing research collaborations, interesting projects, or career opportunities in AI/ML/NLP.
Resume by role