- 🔭 Currently building Autonomous Agent Systems (LangGraph, MCP, CrewAI)
- 🌱 Exploring Fine-tuning LLMs and RAG pipelines
- 👯 Open to collaborate on Transformer-based & AI Agent projects
- 🤝 Looking for opportunities in SWE (Backend/Full-Stack) and AI/ML Engineering
- 👨💻 All projects at github.com/KaushikML
- 💬 Ask me about AI agents, LLMs, Deep Learning, RAG
- 📫 Reach me at roykaushik354@gmail.com
- 📄 My resume: Resume (PDF)
- ⚡ Fun fact: I learn something new every day — and then promptly forget it.
- Languages: Python, C++, C, Java, JavaScript, SQL
- Frameworks & Libraries: TensorFlow, PyTorch, Keras, Scikit-learn, Flask, Django, FastAPI, Gradio
- AI & Agents: OpenAI SDK, LangGraph, CrewAI, MCP Protocol, LangChain, SentenceTransformers, RAG, FAISS, Qdrant, ChromaDB
- LLMs: GPT-4/4o, Gemini, LLaMA, Claude — prompt engineering, tool-use, fine-tuning
- Data & DevOps: Pandas, NumPy, Matplotlib, Power BI, Tableau, AWS, Docker, NGINX, REST APIs, Databricks, PySpark, MLflow, Delta Lake
- Testing & CI/CD: Keploy, PyTest, GitHub Actions, GitLab CI, Jenkins
- Tools: Git, GitHub, Postman, VS Code, Linux, SQLite, PostgreSQL, MySQL, Selenium, NetworkX
- Tech: PySpark, Delta Lake, FAISS, Streamlit, Databricks, MLflow
- Built a Bronze→Silver→Gold data pipeline on Ghana NGO healthcare data
- Hybrid SQL + FAISS retrieval for contextual health queries
- Anomaly detection on patient records with LLM-powered copilot interface
- Role: Backend Lead | Tech: Flask, Google Gemini, RapidFuzz, Serper API, PDF generation
- Secure APIs for file ingestion, AI probability scoring, and citation validation
- Semantic + exact-match checks against live web sources
- Auto-generated PDF reports with originality and AI-content scores
- Role: Full-stack | Tech: Python, SQLite, Gradio, Polygon API, GPT-4, Brave Search
- Multi-agent personas for stock trading simulations with real-time data
- Modular tool-server architecture with tracing, memory tooling, and persistence
- Custom web crawler with intelligent indexing
- Semantic search layer over crawled content using vector embeddings
- Accuracy: 92.6% on cell image classification
- Hybrid CNN + Transformer architecture with custom data augmentation pipeline
- Published comparative analysis on CNN augmentation strategies
- Tech: ChromaDB, SentenceTransformers, Random Forest, GPT-4o Mini, Pushover
- Processed 400K+ product embeddings for smart deal evaluation
- RSS feed scanning every 5 min with push notifications + Gradio UI
- Network routing optimization using NetworkX and QoS metrics
- Flow prioritization with adaptive rerouting under congestion
- 🥈 Nokia Hackathon 2025 — 1st Runner-Up (Plagiarism & AI Content Detector)
- 📜 ServiceNow Certified System Administrator (CSA)
- 🎓 B.Tech CSE, KIIT University — CGPA 8.61
- 📝 Published: Comparative Analysis of CNN Data Augmentation Strategies
- 🧠 Interests: Quantum computing theory, AI ethics, backtracking algorithms, consulting guesstimates
- 🔥 Passion: Building real-world AI systems that automate complex decision workflows
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |