Site Reliability Engineer Intern
Results-driven AI/ML Engineer with production experience in Generative AI, LLMs, RAG, end-to-end ML pipelines, LLMOps, inference optimization, and cloud deployment. Achieved 40% conversion uplift through recommendation systems, 94.2% accuracy in CNN-based medical image classification, and RAGAS scores above 0.87 on agentic AI systems.
Mthree
Jan 2026 – Present · Mumbai
Integrated Gemini LLM for automated AI-driven Root Cause Analysis within Jenkins CI/CD pipelines, reducing MTTR by 35%. Provisioned ML-ready AWS EC2 infrastructure using Terraform and Ansible, orchestrated containerized services on Kubernetes, and established Prometheus, Grafana, and Alertmanager observability for SLA/SLO monitoring and alerting.
Nitorious Atelier
April 2025 – Dec 2025 · Noida
Architected and deployed a collaborative filtering recommendation engine on 5K+ user-item interactions, delivering 40% conversion uplift and 25% cross-sell growth. Engineered 15+ behavioral features, reduced inference latency by 25%, improved marketing effectiveness by 18% through k-means segmentation, conducted A/B tests, and documented model performance and deployment procedures.
B.Tech, Computer Science
2021 – 2025
CGPA: 8.0 / 10. Relevant coursework included Data Structures & Algorithms, OOP, Operating Systems, DBMS, Computer Networks, Data Analysis, Machine Learning, Collaboration, Communication, and Soft Skills.
Built a production-ready agentic RAG system using LangGraph and LangChain with dynamic tool routing between FAISS vector search and Wikipedia API. Developed PDF ingestion, chunking, embeddings, stateful multi-turn workflows, and a Streamlit application, achieving RAGAS faithfulness of 0.91, context recall of 0.87, and answer relevance of 0.89.
Designed and trained an 8-layer CNN on 2,000+ chest X-rays, achieving 94.2% test accuracy and 0.97 AUC-ROC. Applied augmentation, batch normalization, dropout, and learning-rate scheduling, then deployed a Dockerized FastAPI inference service with Grad-CAM explainability.
Built an LSTM transit ridership forecasting model achieving RMSE of 142 versus a Prophet baseline of 198, a 28% improvement. Applied GNNs for congestion-hub node classification with 83% node-level accuracy, tracked experiments with MLflow, and packaged the pipeline with Docker.
Analyzed 2K+ customer records, built Logistic Regression, Random Forest, and XGBoost models, achieved 0.89 AUC-ROC with XGBoost, and reduced false negatives by 18% using SHAP insights and threshold tuning. Developed Power BI dashboards for churn trends and customer segments.