███╗   ███╗ █████╗ ███╗   ██╗     ██╗██╗   ██╗███╗   ██╗ █████╗ ████████╗██╗  ██╗
████╗ ████║██╔══██╗████╗  ██║     ██║██║   ██║████╗  ██║██╔══██╗╚══██╔══╝██║  ██║
██╔████╔██║███████║██╔██╗ ██║     ██║██║   ██║██╔██╗ ██║███████║   ██║   ███████║
██║╚██╔╝██║██╔══██║██║╚██╗██║██   ██║██║   ██║██║╚██╗██║██╔══██║   ██║   ██╔══██║
██║ ╚═╝ ██║██║  ██║██║ ╚████║╚█████╔╝╚██████╔╝██║ ╚████║██║  ██║   ██║   ██║  ██║
╚═╝     ╚═╝╚═╝  ╚═╝╚═╝  ╚═══╝ ╚════╝  ╚═════╝ ╚═╝  ╚═══╝╚═╝  ╚═╝   ╚═╝   ╚═╝  ╚═╝

Final-year ISE student building at the intersection of AI systems, backend engineering, and quality automation.
I work on things that process documents intelligently, classify medical images, route LLM calls, and break software before users do.

focus = ["LLM Systems", "Deep Learning", "Backend APIs", "Agentic QA", "MLOps"]
currently_building = "AI Gateway with TF-IDF routing + LLM guardrails"
open_to = ["AI/ML Engineer", "SWE", "Backend Engineer", "QA Automation", "Agentic Testing"]

Amazon ML Summer School Aug – Oct 2025
→ Supervised/unsupervised learning, deep learning, optimization on 10K+ sample datasets
→ Model evaluation pipelines, hyperparameter tuning, AWS-based ML workflows

Lightweight AI gateway with intelligent request routing and safety enforcement

• TF-IDF based semantic router — classifies and routes prompts to appropriate LLM endpoints without heavy embeddings overhead
• Small neural network classifier (CPU-optimized) — custom-trained DBERT-style model for intent detection and guardrail enforcement
• Embedding-based similarity layer — cosine similarity matching for prompt deduplication and caching
• LLM guardrails — trained lightweight classification model to detect unsafe/off-policy inputs pre-inference

Stack: Python PyTorch TF-IDF Embeddings FastAPI LLM APIs

End-to-end pipeline for extracting structured data from unstructured documents

• OCR preprocessing pipeline for scanned PDF → machine-readable text
• Vision-Language Model (VLM) integration for entity extraction and document understanding
• Structured JSON output pipeline for downstream analytics and automation
• FastAPI backend for document upload, async processing, and scalable extraction

Stack: Python FastAPI VLM APIs OCR SQL AI Automation

Hybrid deep learning system for medical image classification

• CNN + Vision Transformer hybrid architecture for bone marrow cell classification
• Training pipeline with data augmentation and transfer learning — +18–25% generalization improvement
• Evaluation using weighted F1, confusion matrix, and imbalanced-data validation strategies
• FastAPI inference API with latency monitoring and structured logging

Stack: Python PyTorch CNN ViT FastAPI

LLM-powered content validation and misinformation detection

• Gemini API integration with prompt-based workflows for summarization and classification
• Multi-run response logging to assess LLM accuracy, consistency, and reliability
• Structured storage and analytics for research reporting

Stack: Python Gemini API SQL Git

Centralized observability platform for API reliability

• Latency and error-rate monitoring with threshold-based alerting
• Real-time and historical analytics dashboards

Stack: Kotlin MongoDB Next.js

Functional + data-level test automation at scale

• End-to-end UI automation with Selenium + TestNG
• UI ↔ Database consistency validation via SQL
• Agentic test flows — structured defect reporting, regression detection

Stack: Java Selenium TestNG MySQL

Language     →  Python · Java · SQL
AI/ML        →  PyTorch · Transformers · CNNs · ViTs · TF-IDF · Embeddings
LLM Systems  →  LLM APIs · VLM · OCR · RAG · Prompt Engineering · Guardrails
Backend      →  FastAPI · REST APIs · Auth
DevOps       →  Docker · Prometheus · Grafana · Git · CI/CD
QA & Testing →  Selenium · TestNG · Agentic Testing · API Testing · Postman
Data         →  ETL Pipelines · Data Validation · SQL Analytics

B.E. Information Science Engineering
Dayananda Sagar College of Engineering, Bengaluru · 2022 – 2026
Coursework: DBMS · DSA · OOP · Machine Learning · QA · SQL