AI / 2026Smarter intent.
Fewer LLM calls.
An encoder-only transformer for KIOSK intent classification, designed to reduce routine LLM calls and token costs.
Explore the projectI’m Ravishankar. I connect banking expertise with hands-on engineering—building AI solutions, automating workflows, and bringing everyday devices to life.
Independent R&D · Office POCs · A lifelong habit of making things
Three explorations connecting intelligence, software and the physical world.
AI / 2026An encoder-only transformer for KIOSK intent classification, designed to reduce routine LLM calls and token costs.
Explore the project
AI + HARDWARE / 2026Connecting Claude to device control, turning a natural-language request into a physical action.
Explore the project
AUTOMATION / 2024Using Netflix Conductor to bring workflow orchestration into a real-world plant-watering system.
Explore the projectAI, connected devices, weekend builds and experiments. Every project has a story—and a demo.
Designed, architected, implemented and demonstrated an encoder-only transformer for KIOSK intent classification as a cost-saving initiative, aimed at reducing routine LLM calls and token consumption.
Connected Claude to physical device control, demonstrating how a natural-language request can operate a table lamp. An exploration of the bridge between conversational AI and everyday hardware.
Built a voice-driven interface that saves information into a database, exploring hands-free data entry through spoken interaction.
Integrated Alexa with a Raspberry Pi 3 to explore voice control of connected devices.
Built a queue-management prototype with an electronic LED token display and web controls that advance the token number, connecting a browser-based workflow to a physical display.
Updated an LED message display over Wi-Fi using a webpage hosted on an ESP32 operating as an access point.
Used Netflix Conductor workflows to operate plant watering, bringing software orchestration into physical-device automation without third-party smart plugs.
Created a remote plant-watering system triggered by calls to a dedicated phone number, enabling watering while away from home.
Used an SR04 ultrasonic sensor to detect water levels and switch a pump relay at upper and lower thresholds.
Built a water-pump control panel with automatic and manual selection, plus a light indicating the pump’s operating status.
Published temperature readings to a web portal using Node-RED, connecting sensor data to a browser-based monitoring experience.
Read and displayed temperature and atmospheric pressure using a Raspberry Pi and Sense HAT, exploring environmental sensing on a compact computing platform.
Explored sensor-based heart-rate and SpO₂ monitoring and compared displayed readings with a commercial pulse oximeter as a hobby electronics experiment.
Created a watch-style connected display presenting updated COVID-19 case counts, demonstrating a compact interface for online information.
Built and demonstrated remote robot control, including operation over the internet, across a series of working-prototype demonstrations.
Transmitted data from an Android app over Bluetooth, displayed received characters in a terminal and processed the input on an ATmega328.
Explored device control through infrared commands, including a 38 kHz interface using an ATmega328, with demonstrations from 2015 and 2018.
Experimented with an RF transmitter and receiver, observing how receiver output responds to transmitted input.
Explored time-based fan-speed adjustment and Android-operated fan control through two home-automation prototypes.
Built a motion-triggered burglar-alarm prototype using a passive infrared sensor.
Combined an ATmega328, DS1307 real-time clock and 16×2 LCD to display the time and control devices using configured on/off schedules.
Demonstrated an embedded OLED display interface as a proof of concept for compact visual output.
Adapted existing code to create a magic-mirror-style display showing internet-sourced time and date information.
Ran Tetris on an ESP8266 and a TFT screen, exploring embedded graphics and game logic on compact hardware.
Demonstrated a compact Snake game prototype inspired by the classic handheld brick-game experience.
Explored MicroPython running on a Raspberry Pi Pico as a starting point for Python-based microcontroller experiments.
Experimented with low-level VGA video-signal generation using a PIC16F84 microcontroller.
Used the TVOut library with an ATmega328 to generate television-compatible video output.
Demonstrated a miniature servo in operation while exploring small-scale actuator control.
Experimented with a thermoelectric cooling module and demonstrated freezing a small amount of water.
Demonstrated a tabletop setup exploring movement with a slow-speed synchronous motor.
Documented the behaviour of an engine running on a single cylinder as a mechanical observation and experiment.
Try a different technology, category or video format.
I’m a banking technology professional, AI practitioner and technology creator based in South Bangalore. I enjoy taking an idea from the first question through design, architecture, implementation and a working demonstration.
My work spans banking products, AI experiments and SDLC automation. Outside work, the same curiosity leads me to ESP32 boards, Raspberry Pi, embedded displays, robotics and amateur radio. This is where I document and share those explorations.
Designed, architected and implemented AI proofs of concept, including transformer-based intent classification for KIOSK. A hands-on approach to making ideas demonstrable.
Experience across core banking, CASA, payments, loans, trade finance and teller operations, connecting functional requirements with technical implementation.
Designed and developed branch solutions with Codex and Cline. Created Codex skills and agents reused across the Oracle Banking suite of products, and developed tools with Codex CLI at the core to automate daily processes and tasks.
Hands-on exploration of embedded systems, IoT and robotics, shared through working prototypes, videos and Shorts—from microcontrollers and sensors to connected devices.
Open to conversations about AI solutions, banking technology, SDLC automation and ideas worth prototyping.