Data Scientist / ML–AI Engineer in Dublin, building practical AI systems end-to-end — from retrieval and agentic reasoning to deployment and observability. Previously @ Innovaccer. I care about reliability, clear interfaces, and shipping useful tools.
StatsEye — a research assistant for Ireland’s CSO data
Hybrid retrieval + agentic Python analysis on CSO datasets, wrapped in a conversational UI (Chainlit), deployed on GCP (Cloud Run). Makes finding the right table and getting first insights a one-message job.
Designed to transform how researchers, analysts, and curious minds interact with statistical data available on Data-CSO website. Try the StatsEye-Assistant now!
Tools and frameworks used:
- LangGraph (for agent orchestration)
- FAISS Vector DB (for storing vector embeddings)
- Gemini 2.5-flash (API) and Gemini embedding model
- Redis (for checkpointing and long-term memory storage)
- Chainlit (UI-wrapper on React front-end)
- Google Cloud Run (for deployment)
- Docker (for containerising)
Check out my GitHub repo here.
- Languages: Python, SQL
- AI/IR: LangGraph, LangChain, CrewAI, RAG (BM25 + dense + LLM rerank), FAISS
- Apps & Infra: GCP (Cloud Run, Artifact Registry), FastAPI/Flask, Docker, Kubernetes, Redis/Celery, Chainlit, Streamlit
- Data & Viz: Pandas, Seaborn, NumPy, Folium, Jupyter
- StatsEye / Data-CSO-Bot — AI agent over Ireland’s CSO datasets (LangGraph, hybrid retrieval, Python analysis).
Repo:A-Raghav/Data-CSO-Bot
- Multi-Agent LLM Research Assistant — CrewAI-based agents that help draft research sections with structured prompts.
Repo:A-Raghav/Multi-Agent-LLM-Research-Assistant
- FairAI: Ethical Bias Mitigation — Lightweight wrappers over IBM AIF360 + example notebooks for bias detection/mitigation.
Repo:A-Raghav/Ethical-Bias-Mitigation
- Diagnosis-Code Embeddings (GNNs) — Learning ICD-10/HCC embeddings from claims data using GNNs.
Repo:A-Raghav/Diagnosis-Code-Embeddings-using-GNNs
- QueryBridge (Hack-weekends 2023) — Natural-language to SQL transformations to speed up onboarding/ETL rule authoring.
Repo:A-Raghav/hack-weekends-2023
- Data-Zenith (Hack-weekends 2024) — Enterprise search + analytics over org knowledge bases (RAG + agents).
Repo:A-Raghav/hack-weekends-2024
- Flask + Celery on Kubernetes — Minimal worker pipeline with Redis broker/back-end; manifests + compose.
Repo:A-Raghav/Flask-App-with-Celery-Worker-on-Kubernetes
- Market Intel (CMHC vs MaineHealth) — Geo EDA with interactive heatmaps (market share, leakage, proximity).
Repo:A-Raghav/EDA-CMHC-Maine-Market-Intel
- Retrieval that actually retrieves (hybrid ranking, table/entity linking)
- Agentic workflows with safe tool-calling and review gates
- “First-5-minutes” insights: fast EDA, charts, and citations straight from chat
- Clean deployment paths (containers → Cloud Run/K8s) and observability
- LinkedIn: in/aseem-raghav-a65216160
- Email: raghavaseem@gmail.com