recent_projects.html

<h3 class="mb-4">Recent Projects and Initiatives</h3>
<div class="table-responsive">
    <table class="table table-bordered table-striped">
        <thead class="table-light">
            <tr>
                <th>Area</th>
                <th>Details</th>
                <th>Tools</th>
                <th>Status/Duration</th>
            </tr>
        </thead>
        <tbody>


            <tr>
                <td> Agentic AI & Large Language Models</td>
                <td>
                    AI Stock Selector. Agent can analyze the news and predict a sentiment for a given stock.
                </td>
                <td>Langgraph</td>
                <td>In-Progress</td>
            </tr>

            <tr>
                <td> Agentic AI & Large Language Models</td>
                <td>
                    AI Course Creator. Agent will be given textbook and it will be able to create entire
                    course
                    content which can be published in LMS.
                </td>
                <td>OpenAI, Whisper, Langgraph</td>
                <td>2025 (Trials in-progress)</td>
            </tr>

            <tr>
                <td> Agentic AI & Large Language Models</td>
                <td>
                    AI Speech trainer: Personal trainer to improve speech quality. This will be used for
                    giving
                    teaching training.
                </td>
                <td>OpenAI, Wisper</td>
                <td>2025 (Trials in-progress)</td>
            </tr>

            <tr>
                <td> Agentic AI & Large Language Models</td>
                <td>Building a Research Assistant bot which can read locally saved PDF research articles and
                    produce summarised reports
                    <br> In collaboration with <a href="https://sumeetpalsingh.github.io/" target="_blank">Dr.
                        Sumeet Pal Singh </a>, Life Sciences, SNU.
                </td>
                <td>OpenAI, Langchain, Langgraph</td>
                <td>2025 (Trials in-progress)</td>
            </tr>

            <tr>
                <td> Agentic AI & Large Language Models</td>
                <td>AI Grading Assistant: Grading is a very time consuming work for faculty. Late feedback
                    from
                    faculty reduces the chance of improvement. We are building an AI grader, which can match
                    student answer with expected and answer and instantly provide grades to students.
                </td>
                <td>OpenAI, Langchain, Langgraph</td>
                <td>2025 (Trials in-progress)</td>
            </tr>

            <tr>
                <td> Innovative Teaching </td>
                <td>
                    <a href="https://moodle.freebirds.co.in/course/index.php?categoryid=1" target="_blank"> LMS for
                        gamified learning
                    </a>
                    This LMS allows students to records how well they are understanding each class topics.
                    The
                    sequence of topics adjusts according to individual student progress.
                </td>
                <td>Moodle</td>
                <td>Trials in progress </td>
            </tr>

            <tr>
                <td> Machine Learning</td>
                <td>Machine Learning in Stock Market<br>
                    <a target="_blank" href="https://onlinelibrary.wiley.com/doi/abs/10.1002/for.3265">Paper</a>
                </td>
                <td>Keras, Tensorflow, AutoML</td>
                <td>2021- 2025</td>
            </tr>

            <tr>
                <td>Machine Learning</td>
                <td>Automatic Detection of Infected Cells

                    Our biology department routinely takes microscopic images. They use their field knowledge to
                    identify certain patterns to separate out healthy and infected cells. We used object detection
                    techniques to automate this process.
                    <br>
                    <a target="_blank" href="https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=10669553">Paper</a>
                </td>
                <td>YOLO, TensorFlow, Python, Object Detection</td>
                <td>2021- 2024</td>
            </tr>

            <tr>
                <td>Reinforcement Learning</td>
                <td>Intelligent Game agents creation in the Unity Gaming Engine</td>
                <td>MLAgents</td>
                <td>2024 (on hold)</td>
            </tr>
            <tr>
                <td>Reinforcement Learning</td>
                <td>Develop Automatic stocks trader</td>
                <td>Tensortrade, Rllib, TensorFlow, Python</td>
                <td>2022–2024 (on hold)</td>
            </tr>
            <tr>
                <td>Reinforcement Learning</td>
                <td>A RL model to figure out Gym’s Frozen Lake paths<br>Code: <a
                        href="https://github.com/krajit/frozenlake-RL">GitHub Link</a></td>
                <td>Python, Torch, RLlib</td>
                <td>2022</td>
            </tr>
            <!-- <tr>
                <td>AI, Machine Learning</td>
                <td>Face Recognition App for Office Attendance<br>https://github.com/MAT496-Monsoon2025-SNU
                </td>
                <td>Python, JavaScript, HTML, Node.js, Flask, Heroku</td>
                <td>2021</td>
            </tr> -->
            <tr>
                <td>Python Consultancy</td>
                <td>Trained Govt officers in basic Python usage (Cabinet Secretariat, Govt of India)</td>
                <td>Python</td>
                <td>Nov–Dec 2020</td>
            </tr>
            <tr>
                <td>CFD, Auto Diff</td>
                <td>Shape optimization using <em> autodiff </em>:<br>

                    I was trying to use automatic differentiation <a href="http://ceres-solver.org/" target="_blank"
                        rel="noopener noreferrer">http://ceres-solver.org/</a> to solve one shape optimization CFD
                    problem. I noted that this code does not have a handy tool for numerical differentiation. I
                    contributed a small wrapper.
                    <br>
                    Code: <a href="https://github.com/krajit/numericDiffFirstOrder">GitHub Link</a>
                </td>
                <td>C++, CMake</td>
                <td>2019–20</td>
            </tr>
            <tr>
                <td>CFD</td>
                <td>Adaptive Overset Mesh in <a href="https://www.openfoam.com/" target="_blank"
                        rel="noopener noreferrer">OpenFoam</a>:<br>
                    OpenFoam is a popular CFD simulation library. For large deformation fluid structure interactions
                    (FSIs) it provides and overset mesh modules. I combined mesh refinement and overset mesh modules to
                    get the best of both.

                    <br>
                    Code: <a href="https://github.com/krajit/oversetandamrfvmesh">GitHub Link</a>
                </td>
                <td>OpenFOAM, C++</td>
                <td>2019</td>
            </tr>
            <tr>
                <td>CFD</td>
                <td>sixDOF solver in OpenFOAM: Coupling solidBodyMotionFunction with sixDOF solver in <a
                        href="https://www.openfoam.com/"> OpenFOAM </a> <br>


                    OpenFOAM provides a solidBodyMotionFunction class which is convenient for solving flow around moving
                    bodies. However, all the options under this class like, linearMotion, rotatingMotion, etc all seem
                    to solve flow around prescribed motion. I have written a new sixDoFDynamicMotion class, which is a
                    child of solidBodyMotionFunction, which has a private member motion_ of the type
                    sixDoFRigidBodyMotion. The member motion_, is responsible for integrating pressure and shear force
                    around the given patch, and solve for the resulting translation and rotations. The resulting motion
                    is returned as a septernion, through the transform() function. This is tested on two examples.
                    Trajectory of a projectile motion, and a simple pendulum.
                    <br>Code: <a href="https://github.com/krajit/sixDoFDynamicMotion">GitHub
                        Link</a>
                </td>
                <td>OpenFOAM, C++</td>
                <td>2019–20</td>
            </tr>
            <tr>
                <td>Timetabling</td>
                <td>Automated timetable with Unitime<br><a href="https://timetable.snu.edu.in/">Unitime
                        Site</a>

                    <br>

                    Automation of our University timetabling

                    I was the sole administrator, maintainer, trainer of the SNU Unitime instance. This greatly eased
                    the data collection for timetabling as well as automated the timetable generation.

                </td>
                <td>SQL, Unix, SSH</td>
                <td>2019–current</td>
            </tr>
            <tr>
                <td>Timetabling</td>
                <td>Automation of our University timetabling: 
                    Before switching to Unitime, I was the sole timetable developer with the use of the software <a href="https://lalescu.ro/liviu/fet/" target="_blank" rel="noopener noreferrer">FET</a>.
<br>Code: <a
                        href="https://github.com/krajit/csvToFET">csvToFET</a>, <a
                        href="https://github.com/krajit/TT-snu-spring2020">TT-snu-spring2020</a></td>
                <td>Python, FET, XML</td>
                <td>2018–19</td>
            </tr>
            <tr>
                <td>General Purpose App</td>
                <td>Clinic Queue Notification App<br>
                    AN APP TO REDUCE WAITING TIME FOR PATIENTS IN CLINICS/HOSPITALS <br>

I have noticed that whenever we go to see a doctor, the actual consultation takes 5 to 10 minutes, but we have to wait anywhere from 1 to 4 hours in the line. 

I have developed an app which will keep notifying patients about the status of the line. Patients then need not be sitting in the waiting area. The app will tell them what token number is currently with the doctor , and they can come to the clinic only when their number is near. 

                    
                    <br>Code: <a href="https://github.com/krajit/clinicQue">GitHub
                        Link</a></td>
                <td>JavaScript, HTML, Google Cloud, Firebase, Node.js</td>
                <td>2019</td>
            </tr>
            <tr>
                <td>General Purpose App</td>
                <td>Garud Carpool App<br>
                    An experimental app for optimizing cab sharing to my work place Shiv Nadar University and back to my home Greater Noida.
                    <br>Code: <a href="https://github.com/krajit/garud-carpool-app">GitHub
                        Link</a></td>
                <td>JavaScript, HTML, Google Cloud, Firebase, Java, Android</td>
                <td>2018</td>
            </tr>
        </tbody>
    </table>
</div>