I work between the model and the data: getting messy inputs into shape, building the service around a model, and checking the results hold up.
I'm an AI/ML engineer finishing my MS in Information Systems at Northeastern University. I build LLM products and RAG systems from end to end, along with the data pipelines behind them.
- Most recently: AI/ML Engineering Intern at IPSER Labs (Jan β Jun 2026), the main backend engineer on a smart pantry app and its LLM features.
- Before that: Data Engineer Intern at Bharat Electronics Limited and Data Analyst Intern at Compsoft Technologies.
- Certified: AWS Certified Machine Learning Engineer β Associate Β· Salesforce Certified AI Associate
- Next: open to full-time AI/ML, LLM and data engineering roles from January 2027.
|
Oct 2026 Built a medallion pipeline on AWS that pulls live YouTube trending data for 10 regions and refines it into Bronze, Silver and Gold tables with PySpark on Glue. Lambda S3 Glue PySpark Step Functions Athena
|
Dec 2025 Built a multi-agent travel planner on LangGraph and GPT-4o: a root agent routes each request to 1 of 3 specialist agents, by keyword rules first and an LLM as fallback. React FastAPI LangGraph GPT-4o FastMCP
|
|
Oct 2026 Β· try the app β Built a phone-first vault for prescriptions, reports and hospital bills: Gemini and open models on Groq read each page, and every record is badged with its source. Streamlit Gemini Groq Turso CMS APIs
|
Aug 2025 Moved 1M+ options records from AWS S3 through a 6-layer Snowflake medallion architecture in under 5 minutes, cutting market-data preparation time by 70%. AWS S3 Snowflake UDFs Cortex agents LLM agents Streamlit + Slack
|
|
Dec 2025 Built a financial RAG chatbot that sends structured questions to DuckDB and document questions to Chroma, with document-level access control across 6 user roles. FastAPI DuckDB Chroma GPT-4o Cohere
|
Dec 2025 Built a real-time lakehouse pipeline that processed 10M+ transit records at 1,000+ events per second with sub-5-second latency for live service monitoring. Kafka Spark dbt BigQuery Power BI
|
| Project | What it is |
|---|---|
| MindNet Β· May 2025 | A multi-task mental health conversational model that runs locally, trained on 25,000+ labeled interactions: 88% accuracy and 0.84 macro-F1 across intent and emotion. |
| Graph-Based EEG Seizure Detection Β· Feb 2025 | 1,000+ EEG recordings from 200+ patients turned into graph features that capture how electrodes relate in space and how seizure patterns change over time. |
| Smart Hydroponic Farm Management Β· Dec 2024 | A normalized Oracle database for 3 simulated farms tracking 50,000+ crop, nutrient, environmental, inventory and harvest transactions. |
| AbDiffuser | A diffusion pipeline that generates antibody sequences and 3D structures together, reproducing the APMixer architecture and its SE(3)-equivariant design. |
All ten projects, with drawings of how each one works, are on my portfolio.
- Evaluating a RAG system when access control matters: what FinSight taught me about routing questions to the right store, enforcing permissions at retrieval time, and why leakage is a count, not a percentage.
- Making an LLM layer provider-agnostic: how I put Claude in production and Ollama in development behind one interface at IPSER Labs, with structured outputs as the contract.
- A mental-health chatbot that knows when not to generate: designing MindNet so crisis messages never reach the text generator.
AI / ML PyTorch TensorFlow scikit-learn Hugging Face Transformers Deep Learning NLP Feature Engineering Model Evaluation
GenAI / Agents LLMs RAG LangChain Agentic AI Prompt Engineering Claude API Ollama GPT-4o Cohere MCP Structured Outputs Guardrails
Data / Databases SQL Oracle / PL/SQL Kafka PostgreSQL MySQL MongoDB Snowflake BigQuery DuckDB Spark / PySpark Airflow dbt Pandas
Cloud / Backend Python FastAPI REST APIs AWS S3 AWS Textract AWS Lambda AWS Glue Step Functions Athena Docker Git
Languages & Web Python Java SQL TypeScript React HTML / CSS
Viz & Tools Power BI Tableau Streamlit Matplotlib Jira Postman Jenkins
Thanks for stopping by. Say hi to Guru on my portfolio: press / anywhere there to ask it something.


