Pratham Kumar
Education
Manipal University, Jaipur
Bachelor of Science in Computer Science and Engineering
Expected Graduation: 06/2024
Relevant Coursework:
- Data Structures and Algorithms
- Software Engineering
- Data Science and Machine Learning
- Computer Networks
- Computational Neuroscience
- Improving Deep Neural Networks
- Structuring Machine Learning Projects
- Blockchain
- Neural Networks and Deep Learning
- Summer Academic Internship in Deep Learning at NUS, Singapore in partner with Hewlett Packard
Technical Skills
Programming Languages: Python, GO, C++/C
Frameworks/Libraries: Flask, Django, TensorFlow, OpenAI, LLAMA, LangChain, LlamaIndex, Redis, Celery, Streamlit, React, Node, NextJS, ExpressJS, vLLM, DDIMs, stable diffusion, Pydantic
Tools/Technologies: Git, Docker, Amazon S3, Amazon EC2, AWS ECS, AWS Lambda, Amazon DynamoDB, Google Cloud Platform, SQL, MongoDB, Heroku, Render, Postman, Automations, Project Management, LLMs
Projects
Baani (JuggernotAI)
- An LLM-based application capable of creating, scheduling, and posting content for businesses on social media platforms
- Created the entire backend using Flask, fine-tuned GPT-3.5-Turbo on different client data, minimized response latency, implemented function calling for sending and creating emails
- Worked with Instagram and Twitter’s API, added database memory to chat LLM, and created an agent network for self-iteration over feedback
ReplyRocket (JuggernotAI)
- An automation agent that drafts and posts replies to content made by authors in a tracking list, helping businesses scale their social media influence
- Created backend using Flask, managed APIs for handling server-side queries, implemented monolithic architecture and Docker container
- Worked with Twitter APIs, rate limits, efficient information generation and retrieval, authentication APIs (OAuth1.0, 2.0(PKCE), JWT)
- Created an assistant using GPT-4 API with retrieval, function calling, and custom instructions
- Worked with Pydantic dataclasses and validators for creating user classes, created a real-time tweet retrieval algorithm using a background cron worker in GO
- Converted 15+ users to purchase the product, 50+ users to try the product, achieved 100% customer satisfaction and app uptime rate
Serveza
- An app that connects people interested in doing small gigs for managing everyday tasks in real-time
- Created the entire app with a React frontend, Flask backend, and Docker containerization
- Implemented a real-time device tracking engine using GPS tagging, fine-tuned LLM to understand the nature of gigs for custom pricing
- Empowered the local community by creating jobs, fostered an open culture for freelance services, and developed a mobile app using Apache Cordova
Imagre - Reconstructing Visual Images from fMRI data using DCGAN
- Combined a reconstruction network and GAN to recreate visual images from fMRI data
- Pre-processed fMRI data from ILSVRC2012 and fMRI on ImageNet (Fall 2011 release) of the visual cortex (VC) region
- Created a DCNN and GAN for reconstruction
Work Experience
JuggernotAI - Automation and Machine Learning Engineer (09/2023 - 03/2023)
- Created scalable architectures and system designs for multiple automation agents and backend services
- Developed the backend for ReplyRocket and Baani, landed the first and 15+ clients for Juggernot
- Achieved customer satisfaction from all customers, improved backend architecture and cost-efficiency
- Worked on user security, data protection, and integrity, created secure information exchange systems
- Worked with cloud services in AWS and GCP, created access patterns, blueprint designs, PRDs, user journeys, lean canvases, and documentations
- Created a neural network-based Track degradation model for Continuous-time health analytics of track geometry
- Pre-processed line datasets for multiple up-down lines in Indian Railways, created a Neural Network for determining GMT given different track metrics
- Optimized model training by training sub-models with multi-threading, created a Docker container for project delivery to the server
Ekatvam Innovations Pvt. Ltd (A TCS DISQ company) - Data Analyst (08/2021 - 03/2022)
- Worked on the core backend algorithm, shifted the entire platform to AWS from Heroku
- Developed the dashboard using ReactJS, NodeJS, ExpressJS, Python, Flask, and PostgreSQL
- Worked with PostGIS system, mathematical modeling, and budgeting
- Utilized AWS services (EC2, Route53, Lambda, CodePipeline, Amplify) and Heroku for DevOps
- Worked with Google Maps API and GIS mapping APIs
Honors/Awards
- Gold Medalist - National Cyber Olympiad (2018)
Activities/Involvement
- Head of Projects - Randomize (The official Computer Science club of Manipal University, Jaipur)
- Received honor from India Development Foundation for successfully conducting the resource mobilization for humanitarian causes program