Rakesh Surampalli

Master of Science, Business Analytics

University of Texas at Dallas · 2021 — 2023

Bachelor of Technology, Mechatronics Engineering

Manipal Institute of Technology · 2016 — 2020

Project Manas

MIT Manipal's autonomous-vehicle research team

• Student-run team building autonomous ground vehicles. projectmanas.in ↗

The problem

• Retail investors can see what they hold, but not what it means. Concentration risk, drift against a benchmark, and quietly underperforming positions all stay invisible until they hurt.

What I built

• A conversational analyst: five specialist agents share one thread, so a question about risk, performance, or a single holding lands with whichever specialist should answer it.
• Brokerage linking, so analysis runs on real holdings rather than a manually typed watchlist.
• Automatic surfacing of risk, concentration and opportunities, benchmarked against the S&P 500.

How it is put together

• Python services behind versioned REST endpoints, containerized with Docker and hosted on Azure App Services.
• PostgreSQL for holdings, positions and computed analytics.
• Market data pulled on a schedule and cached, so the conversational layer reads from a warm store instead of hitting the provider per question.

PythonPostgreSQLAzureDockerLLM agentsMarket data APIs

Outcome

• Add the numbers only you have — users, holdings analysed, latency, anything you measured. A specific figure here is worth more than a paragraph.

The problem

• To-do apps hold a list but ignore geography. Errands get scheduled in an order that has you crossing the city twice, and the app never notices.

What I built

• A planner that reads your tasks and orders the day around where you actually are, not just when things are due.
• Shopping runs grouped by location, with price comparison so a trip is worth making.
• Routines and recurring plans, so the schedule reshapes itself rather than being rebuilt each week.

How it is put together

• React Native client sharing components with the web build.
• Python backend for task parsing and scheduling, on Azure with PostgreSQL.
• Maps and place data for the location layer; suggestions are generated server-side so the phone stays cheap to run.

React NativePythonAzurePostgreSQLDockerMaps APIs

Outcome

• Fill in what you measured — retention, tasks planned, time saved per trip.

The problem

• Between a farmer and a buyer sit several intermediaries. Each one takes margin and adds latency, and the farmer — who carries all the growing risk — keeps the least of the price.

What I built

• A marketplace putting FPOs and individual farmers directly in front of buyer demand.
• Crop analytics and IoT soil sensing feeding recommendations on what to plant and when to sell.
• Weather alerts and irrigation signals, so the advice is operational rather than only commercial.

How it is put together

• JavaScript front end with a Python analytics service behind it.
• Firebase for auth, realtime listings and device telemetry — chosen because rural connectivity is intermittent and the offline story mattered more than query flexibility.
• Mobile-first throughout: most users arrive on a phone, on a slow connection.

JavaScriptPythonFirebaseIoT sensorsAndroid / iOS

Outcome

• Add real figures — farmers onboarded, FPOs, volume traded, margin returned to growers.