Hi, I'm Ruchika 👋
Data scientist and AI engineer with a background in Computer Engineering and an M.S. in Information Systems from UT Austin. I bring together strong technical skills and a problem-solving mindset — building predictive models, streamlining workflows, and uncovering insights that lead to action.
Throughout my journey, I've had the opportunity to work, contribute, and collaborate with these organizations.
About me
I started in Computer Engineering at Mumbai University and recently completed my Master's in Information Systems with a focus on Data Science and Analytics at UT Austin.
I've always been fascinated by how data tells stories. I enjoy uncovering patterns, building analytics solutions, and translating complex datasets into insights that drive smarter decisions. From SQL and Python to machine learning and AI, I love building systems that make data useful.
Lately, I've been diving deeper into agentic AI, multi-agent workflows, LangGraph, CrewAI, and RAG to explore how intelligent systems can automate decision-making and augment human workflows.
Curious? Ask my AI career assistant anything about my projects, experience, or technical work.
How I work
Good decisions begin with evidence, not assumptions.
Finding patterns across messy data is where the fun begins.
Ship solutions, learn from feedback, and iterate.
Every dashboard, model, or pipeline should solve a real problem.
Portfolio
13 projects across autonomous agents, LLM systems, ML, and analytics — each with code, write-up, or live demo.
An autonomous AI co-worker built on LangGraph. Give it a goal, tell it what success looks like — it picks tools, does the work, evaluates itself, and keeps iterating until the job is done.
Impact — Worker + Evaluator loop, six real-world tools (browser, Python, files, search, Wikipedia, push). Delivered a research report on Phoenix pizza autonomously — no follow-up prompts.
A CrewAI-orchestrated engineering team — Eng Lead, Backend, Frontend, QA — that designs, builds, and validates full applications collaboratively.
Impact — Hierarchical agent workflow that decomposes a product brief into specs, code, and tests. Demonstrates end-to-end autonomous software delivery.
Production-style investment research crew that discovers, analyzes, compares, and recommends high-potential stocks through hierarchical agent orchestration.
Impact — RAG-backed memory + external market APIs. Agents debate, rank, and produce an underwriter-grade recommendation autonomously.
Autonomous multi-agent research workflow that plans search strategies, executes parallel web research, and synthesizes structured cited reports.
Impact — Async agent orchestration with tool-calling. Turns a fuzzy research question into a structured, cited report — no analyst in the loop.
An autonomous sales development pipeline: specialized agents draft, evaluate, style, and dispatch cold emails through SendGrid.
Impact — Three writing agents compete; a manager agent picks the winner, formats HTML, and ships the send. A lightweight AI sales org in one workflow.
A conversational portfolio agent that answers real-time questions about my projects, skills, and research — with tool-calling and context injection.
Impact — Deployed on Hugging Face Spaces with DeepSeek. End-to-end LLM app architecture: prompt engineering, orchestration, deployment.
Forecasts monthly U.S. retail sales by fusing internal history with macro indicators — CPI, unemployment — across linear, tree, and LSTM models.
Impact — Benchmarked Linear Regression, Random Forest, XGBoost, and an LSTM. Deep net captured non-linear seasonality classic models missed.
End-to-end churn model on telecom customer data with a Tableau risk-intelligence dashboard for the retention team.
Impact — Predictive model + business-facing dashboard highlighting high-risk cohorts — bridging model output and retention action.
Connects public search interest (Google Trends) to weekly ad performance to explain — and forecast — ROAS movement.
Impact — Full pipeline: Trends ingest → weekly aggregation → attribution model → Tableau story. Guides where marketing spend earns its keep.
150,000+ EV records analyzed across U.S. states: BEV vs PHEV, range trends, top models — interactive Tableau storytelling.
Impact — Reveals how policy, price, and range converge to shape adoption.
Classification pipeline on the Pima dataset: EDA, RFE feature selection, SMOTE balancing, model bake-off, Streamlit deployment.
Impact — A production-shaped ML workflow from raw data to a clinician-facing screening interface.
Comparative study of RNN, GRU, and LSTM architectures for forecasting Air Quality Index across urban monitoring stations.
Impact — Investigates trade-offs between architecture complexity and forecast accuracy on real-world environmental sensor data.
Testimonials
Here's what some of them have to say.
Credentials
On Social
I write about what I'm learning in public.
Notes on data science, AI, analytics, and lessons from my journey.
View LinkedIn →