ML engineer focused on building production-grade systems — multi-agent pipelines, LLM-powered apps, and MLOps infrastructure that actually ships.
MS in Data Science at Northeastern (GPA: 3.81) · Graduating Dec 2026 · Boston, MA
- Multi-Agent AutoML — LangGraph system with 4 specialized subgraphs (classification, regression, clustering, feature engineering), supervisor + planner + memory agents, MCP-exposed tooling. Reduced project turnaround from 1 week to 3 hours. Used by real clients.
- RAG Pipeline on GCP — Gemini 1.0 Pro + Pinecone hybrid retrieval, 71% → 86% answer relevance (RAGAS), deployed on Cloud Run at 500+ daily queries.
- MLOps Demand Forecasting — XGBoost + Prophet with Optuna tuning, Airflow orchestration, DVC + GCS versioning, automatic rollback on regression. Full GCP deployment with MCP chatbot endpoint. Demoed at Google Cambridge.
- Knowledge Graph Builder — LoRA fine-tuned LLaMA, GGUF + Llama.cpp local inference, Neo4j graph querying.
- Machine Learning & GenAI Intern — Seagate Technology
- ML Intern — Transformly AI, Boston (LangGraph multi-agent AutoML, MCP)
- ML Intern — Fuzzy Cloud Inc. (RAG on GCP, Vertex AI, Gemini)
Contributor to traceloop/openllmetry (LLM observability SDK) and abetlen/llama-cpp-python:
- 🟢 [Merged] Fixed model metadata parsing in llama-cpp-python → PR #2217
- ✅ [Approved, pending merge Fixed semconv-ai version bump for
GenAICustomOperationNamein openllmetry → PR #3826
LLM & Agents: LangGraph, LangChain, LlamaIndex, MCP, Pinecone, Vertex AI, Hugging Face
ML & MLOps: PyTorch, XGBoost, Scikit-learn, MLflow, Optuna, Airflow, DVC
Deployment: FastAPI, Docker, Cloud Run, GCS, Cloud SQL, GitHub Actions
Languages: Python, SQL, C++, Go
- 🏆 Best Use of Public API — MLH Hackathon (1,614 participants): fine-tuned Xception for real-time pothole classification
- 🏆 Best All-Girls Team — MLH Hackathon (501 participants): pose classifier using TF CNN across 7 pose classes
Reinforcement learning, model optimization, and agentic memory architectures.