Hey, I'm Sujit — a B.Tech Computer Engineering student at VIT Pune who builds production-grade AI systems, secures LLMs, and ships real-world cloud infrastructure.
I work at the intersection of AI Security, Cloud-Native Systems, and Generative AI — obsessed with building systems that are not just accurate, but robust, explainable, and scalable.
"Don't just train a model. Deploy one that survives the real world."
🎓 B.Tech Computer Engineering @ VIT Pune | Graduating 2028 📍 Pune, Maharashtra, India 📮 sujitpatil2006@gmail.com 🎯 Open to: Full-Stack · ML/AI · Generative AI Internships
⚡ More about how I work
- 🏗️ I ship end-to-end: data pipelines → model → serverless infra → monitoring
- 🔒 I think adversarially — every LLM app I build gets red-teamed first
- 📐 I benchmark everything: NSL-KDD, CICIDS2017, UNSW-NB15, custom eval pipelines
- 🌍 I build for real users — JanSevaAI serves live civic use cases on AWS
|
🤖 LLM Agents & RAG Architectures Retrieval optimization, cross-encoder reranking, vector databases, eval pipelines ☁️ AI Infrastructure on AWS Bedrock, SageMaker, Lambda-first serverless design |
🛡️ LLM & API Security Prompt-injection defense, anomaly detection, adversarial robustness 🧩 DSA & System Design 700+ LeetCode solved · production-level engineering patterns |
| Title | Venue | Year | |
|---|---|---|---|
 |
AI-Driven Anomaly Detection for GraphQL APIs under Nested Query Attacks | arXiv (co-authored) | 2025 |
| Project | Tech Stack | Impact | |
|---|---|---|---|
| ⚡ | JanSevaAI — Civic AI Platform | React 19 · AWS Lambda · Bedrock · YOLOv8 · DynamoDB · S3 · CloudFront | Live on AWS · 13 Lambda functions · 3 roles · SNS OTP |
| 🛡️ | LLM Prompt Injection Firewall | Python · ML Classification · Rule-based Filtering | 85% detection accuracy · real-time sanitization |
| 🔍 | GraphQL Anomaly Detection | LSTM · CNN · TensorFlow · OWASP tools | 92% accuracy · arXiv published |
| 🧠 | CodeSense RAG — Codebase Q&A | Python · Embeddings · Vector Search · Cross-Encoder Reranking · CI/CD | Production-grade RAG · full eval pipeline |
| 🏗️ | Multimodal RAG — Construction Safety | PyTorch · SageMaker · AWS Lambda · S3 · Vector DB | Serverless · production SageMaker endpoints |
| 🤖 | AI Network Intrusion Detection | XGBoost · Autoencoders · LSTM · NSL-KDD · CICIDS2017 | Hybrid ML + DL · UNSW-NB15 benchmarked |
|
👨💻 Languages |
🎨 Web · Full-Stack
|
|
🤖 AI · ML · Data
|
|
|
✨ Generative AI
|
☁️ Cloud & Infrastructure
|
|
🗄️ Databases |
🛠️ Tools |
| Achievement | |
|---|---|
| 🧩 | 700+ LeetCode problems solved — strong DSA & problem-solving foundation |
| 📄 | Co-authored arXiv paper on AI-driven GraphQL anomaly detection |
| ☁️ | Live AWS deployment — JanSevaAI serving real civic use cases |
| 🔒 | 85% LLM attack detection — prompt injection firewall in production |
| 🏆 | Kaggle competitor — NVIDIA Nemotron Model Reasoning Challenge (LoRA fine-tuning) |
| 🌍 | Open-source contributor — multiple AI/ML & full-stack projects shipped |
