Prakhar Jain

Prakhar Jain

AI Engineer | Roboticist | FPV Drone Pilot

About Me

I am an engineer very passionate about AI, code and robotics. My journey started as a Mechanical Engineer from IIT Bombay, where I worked to build an electric car to participate in Formula Student UK Competition. From there, I moved on to explore robotics and did my master's thesis in autonomous drones. I worked at Jaguar Land Rover for 2 years, focusing on thermal modelling of electric motors.

Fascinated by AI in healthcare, I moved to Qure.ai. I first started working on Qure.ai's flagship product qXR which is AI fast reporting in Chest X-Rays. Currently, I am working on building Qure.ai's Stroke Suite which is an AI powered solution for early triaging and timely treatment of stroke patients.

I am extremely fascinated by drones. In my free time, I build and race FPV drones. I love to tell stories with my drone videos as you will see below.

Solving Life with Code

Senior AI Engineer | Stroke Suite

Qure.ai | Bangalore | Oct 2023 - Present

  • Developed end to end pipeline to process Head CT scans and generate outputs consumable by clinicians
  • Supported highly custom and stable onpremise deployment at client's servers | Trained SREs to support deployment at sites
  • Building demos setup showcasing new features at conferences like World Stroke Congress | Demoed Stroke Suite at ESMINT 2024 to neurointerventionists
  • Ensured pipeline stability by implementing comprehensive Grafana, Prometheus, and Datadog-based monitoring systems, achieving <1% failure rate, 5-minute turnaround time, and 100% uptime.

AI Engineer | qXR

Qure.ai | Bangalore | July 2022 - Oct 2023

  • Developed a gateway to enable hospital modality like Xray, CT machines to send acquired images to qureapp backend
  • Developed an end to end pipeline for qXR from acquisition to AI analysis to output generation and displaying outputs on PACs viewer or Qureapp
  • Worked on qXR v3 models leading to a turn around time improvement from 300 seconds to 20 seconds per scan with high throughput
  • Ensured stability of the pipeline by ensuring code coverage > 80% with automated CI/CD testing, undertaking benchmarking for server processing optimization

Mobius | LLM powered internal tooling

Qure.ai | Dec 2024

  • Created a single place with access to all code, slack, jira, wiki capable of adding context to the conversation in runtime using openai function calling
  • Built a feature library on top of it like AI dashboard using connection to read replica, ability to fetch code context with natural language, ability to answer questions with correct context with a large knowledge base
  • Also gives sources of information to the user to verify the answer
  • TraceMyJourney | Chrome Extension

    Chrome Built in AI Hackathon | Nov 2024

  • Developed a Chrome extension that visualizes browsing history through interactive word clouds, enabling users to filter their online activities with ease. This tool leverages AI to generate summaries, poems, and jokes based on user history, enhancing self-awareness and productivity.
  • Implemented real-time tagging and summarization using Prompt API and Summarization API, allowing users to quickly access and revisit their browsing history. The extension continuously updates, providing dynamic and personalized insights into user behavior.
  • Link to devpost about it - Trace My Journey
  • Natural Language SQL Query Generator

    Hobby Project | Qure.ai | Aug 2024 - Present

  • A sql query generator which takes in natural language as input and converts it to sql query, useful for non technical users to interact with databases. Building this for enterprises for very large and complex database architecture
  • Implementing this local models like llama 3.1, RAG on database schema, telemetry data and existing dashboard queries. Used crew.ai framework to simplify interaction with llm model
  • Meal Planner

    Hobby Project | July 2024

  • Created a template for weekly meal planning, enabling users to set their entire week's meals in under 2 minutes.
  • Includes a chatbot which asks user's preferences, demographics and food preferences and generates a personalised full week meal plan by using this information as prompts to openai models - Weekly Meal Planner
  • Undercover Agent

    Hobby Project | March 2024

    • Free version of the game undercover agent with nearly 4000 words supporting the roles of Mr White, Civilians and Undercover Agent | Link to the game - Undercover Agent

    Passion for FPV Drones

    I am extremely passionate about FPV drones. I love their speed, agility and the rush of adrenaline that they give. I have built 4 drones so far which I use for racing and aerial photography. I got into FPV since 2021 and haven't stopped since then.

    FPV Drones

    My collection of custom-built FPV drones ready for action

    Drone Racing

    Exhilarating moments from an FPV drone race

    I often get asked how I build these drones. It generally comprises of buying each component, putting it together, soldering components, flashing the firmware and configuring the settings. After building the drone, the harder part is learning to fly and control them. These are very fast and agile and I have spent 100s of hours on a simulator to get better at them. I have participated in multiple races too. In the 2022 IDRL race held at IIT Delhi, I was able to secure 3rd position.

    When you fly drones, very far away from you, it is very stressful. You have to learn to deal with the risk, fly the drone and get it back. You have to be ready for it to crash and only then you can improve as a great pilot. All in all, it is super fun to fly them. When you record your video, edit it, add a song on it, it all comes together and gives an amazing memory. When I travel, I always make it a point to make cinematic videos of the place I am visiting. Here are a few of my favourite videos:

    Life as a Mechanical Engineer

    Generation of Impulsive Forces by a Quadcopter through a Cable Suspended Mass

    INDUS Lab, IIT Bombay | Jan 2019 - July 2020

    • Simulated collision of a suspended mass on a quadcopter, with a stationary mass; to drive the latter into a hoop
    • Presented a set of reasonable assumptions on desired trajectories such that the trajectory optimization problem only has linear constraints, to make it computationally efficient to solve
    • Published in the 2020 IEEE/ASME International Conference on Advanced Intelligent Mechatronics

    A Novel Four Degree of Freedom Foot Interface for Controlling a Robotic Arm

    RoMI Lab, Monash University | May 2018 - Dec 2018

    • Designed a foot interface which maps the movement of foot in 4 directions to the motion of a slave robotic arm
    • The interface is designed to allow surgeons to control the position of the endoscope with their foot whilst performing the surgery with their hands, thus making them fully autonomous in performing laparoscopic surgery
    • Published in the 2020 IEEE RAS/EMBS International Conference for Biomedical Robotics and Biomechatronics

    Graduate Mechanical Engineer

    JLR | Bangalore | Sept 2020 - July 2022

    • Developed mathematical models in Ansys and Simulink to predict the performance of electric motors under different duty cycles.
    • Trained LSTM model on experimental PMSM and IC motor temperature data achieving an accuracy of 5°C.

    Steering and Vehicular Dynamics

    Design Engineer, IIT Bombay Racing Team

    • Designed, fabricated and assembled components of steering system of the car reducing the weight by 60%
    • Did tyre selection, behicle dynamics modelling for suspension planning and weight distribution of the vehicle
    • The team won Formula Student award for fourth consecutive time and stood 4th best in Engineering Design among electric vehicles in Formula Student UK'18,

    Research Papers

    Scholastic Achievements