MS Computer Science (AI) @ Purdue · AI&ML Research · LLMs · Cloud & Data Engineering

I'm an MS Computer Science (AI) student at Purdue University, working as an AI&ML Research Assistant across two labs — CIVS and NEXIS. I specialize in building intelligent systems that bridge research and production: from EEG-to-language decoding pipelines to cloud-native NLQ-to-SQL engines.

• 🔬 AI&ML Research Assistant at CIVS (blast furnace optimization) & NEXIS Lab (Brain-Computer Interfaces)
• 🧬 Building EEG → Natural Language decoding pipelines using LLMs and encoder-decoder architectures
• ☁️ 3x AWS Certified — ML Engineer, Data Engineer, Cloud Practitioner
• 🚀 Previously: Data Analyst @ LatentView Analytics — contributed to $1.1M revenue uplift
• 📍 Greater Chicago Area | Open to AI/ML Engineering roles

AI & LLMs: PyTorch, HuggingFace Transformers, LLaMA, GPT, LoRA/QLoRA Fine-tuning, RAG, LLM Agents, Prompt Engineering, JAX

ML & Data Science: XGBoost, Time Series Forecasting, Feature Engineering, Signal Processing, A/B Testing

Vector DBs & Retrieval: Pinecone, ChromaDB, FAISS, LangChain, Semantic Search

MLOps & Infra: Docker, AWS SageMaker, MLflow, Weights & Biases, CI/CD, n8n

Cloud & Data Eng: AWS (Lambda, Glue, Kinesis, Redshift, S3, EC2), ETL Pipelines, Apache Spark

• Built silicon content prediction models for blast furnace optimization using XGBoost — MAE of 0.03% on industrial sensor data
• Engineered time series forecasting pipelines on multivariate sensor streams, improving accuracy by ~22% over baseline ARIMA
• Processed 10K+ timestamped sensor readings with feature engineering and anomaly filtering

• Developed a Brain-Computer Interface (BCI) pipeline decoding EEG signals into natural language using LLMs — BLEU score of 0.42
• Designed MLM-based masking on EEG token sequences, reducing signal artifact noise by ~31%
• Built encoder-decoder architectures aligning EEG embeddings with LLM representation spaces, cutting cross-modal alignment loss by 28%
• Processed and curated 50K+ EEG trial segments for robust cross-subject generalization

• Contributed to a $1.1M revenue uplift via a GenAI-powered virtual assistant (LLaMA) reducing cart abandonment by 74%
• Segmented 11K+ users with K-Means on AWS SageMaker for targeted personalization
• Built real-time analytics pipelines reducing data latency from 30s → milliseconds
• Analyzed 1M+ zero-result search queries using AWS Glue and Athena

• Designed auto-scaling cloud infrastructure with EC2, Auto Scaling, and ELB
• Reduced peak-time latency by 30% through optimized scaling strategies

Hybrid NLQ-to-SQL system combining LLM inference (Groq + LLaMA) with template matching over 100K+ EHR records

• ~95% SQL generation accuracy with sub-500ms response times
• Cloud-native deployment: Docker, AWS EC2/S3, Neon PostgreSQL, FastAPI + React dashboard
• ETL pipeline for FHIR-compliant Synthea healthcare data with audit logging

Knowledge Distillation GAN for 4x image super-resolution — ~94% SSIM on benchmarks

• Compressed student generator to ~40% fewer parameters than baseline SRGAN
• Perceptual loss combining VGG feature loss and adversarial loss for texture preservation

🎓 MS Computer Science (AI) — Purdue University, Hammond, Indiana (Aug 2025 – Present)

Machine Learning · Deep Learning · Data Mining · Big Data Systems · Database Systems

🎓 B.Tech CSE (AI) — Amrita Vishwa Vidyapeetham, India (Jul 2020 – Jun 2024)

NLP · Deep Learning · Reinforcement Learning · Cloud Computing · AI in Speech Processing

"From EEG signals to SQL engines — I build AI that bridges research and the real world."