Open to opportunities

Hi, I'm Ravi Yadav 👋

Building Intelligent, Scalable Systems | AI/ML Researcher | IEEE APPEEC 2026 Co-Author

Ravi Yadav portrait
Machine LearningDeep LearningNLPComputer VisionTime Series ForecastingScikit-learnTensorFlowPyTorchLLMsRAGLangChainLangGraphOpenAI APIPythonSQLJavaScriptTypeScriptJavaOOPNumPyPandas
MatplotlibPostgreSQLMySQLMongoDBSupabaseFirebasePrisma ORMReactNext.jsNode.jsExpressREST APIsSystem DesignGit/GitHubTailwind CSSHTMLCSSSwaggerUI/UXData Structures & AlgorithmsProblem-solving
01Introduction

About

I'm an AI undergraduate (B.Tech, 2023–2027) and AI/ML researcher interested in a simple question: how do we build intelligent systems that work reliably in the real world, not just on a dataset?

My current research explores deep learning for energy and physical systems, where models need to respect the underlying laws of the system rather than simply fit patterns in data. I'm the second of five authors on a physics-informed load forecasting paper presented at IEEE APPEEC 2026 in Singapore, as part of my ongoing work on power-grid forecasting and optimization.

Beyond ML research, I enjoy understanding how systems are built from the ground up — from backend engineering and APIs to high-level design, system architecture, scalability, and reliability. I'm also deepening my foundations in NLP and computer vision, while competitive programming keeps my algorithmic and mathematical thinking sharp.

Long term, I'm interested in graduate research at the intersection of machine learning, physical systems, and infrastructure, while continuing to build scalable software along the way. Always happy to connect over research ideas, interesting systems, or opportunities to build something meaningful.

02Experience

Internships

  • Algocept

    Software Engineer Intern

    January 2025 – April 2025 · 4 mos

    Noida, Uttar Pradesh, India · Remote

    • Improved a React.js and Tailwind CSS admin dashboard by resolving 20+ UI/UX and responsiveness issues, delivering a more consistent cross-device experience.
    • Developed a RESTful API supporting 3 country-specific configurations using NestJS, TypeScript, and MongoDB, enabling dynamic footer content.
    • Optimized admin dashboard loading for 500+ users by implementing pagination, reducing initial load time by ~50% while limiting each request to 5 users.
03Research

Research Publications

  • KAT-PatchTST: Physics-Informed Forecasting with Kirchhoff Conservation

    C. Murali Madhav, Ravi Yadav, K. Mehra, A. Tewary, S. Aggarwal

    18th Asia Pacific Power and Energy Engineering Conference (APPEEC), Singapore · August 2026 · IEEE · Paper ID 190
    • Accepted and presented at IEEE APPEEC 2026; conference proceedings publication pending.
    • Stage 1 of the Watt-IF research project; Stages 2 and 3 under active development, targeting ICML.
    • Provisional patent filed.
    Watt-IF on GitHub
04Work

Check out my latest work

Featured research

Watt-IF – Electricity Data Mining and Grid Resilience Research

  • KAT-PatchTST (Stage 1, 2nd author): Developed a physics-informed BA-aggregate load forecasting framework combining Channel-Independent PatchTST, TimeXer cross-attention, Kirchhoff conservation regularization, and ReLoBRaLo dynamic loss balancing. Achieved 3.55% demand MAPE across six BAs on the EIA930 protocol using a 168-hour context window (30% shorter than the published 240-hour baseline) and approximately 0.6M parameters, making the model 3–10× leaner than comparators.
  • Stage 2 (In Development): Extending the forecasting framework toward operational feasibility analysis over the BA interchange network to identify systemic bottlenecks and quantify node criticality from forecasted grid states.
  • Stage 3 (In Development): Developing a conditional grid partitioning policy to minimize allocation failures under forecasted operating conditions, motivated by ORNL’s reported $121B annual cost of major U.S. power outages in 2024.
PythonXGBoostDeep LearningTensorFlowData MiningGraph Theory

PrepLens – Placement Prep Platform

  • Built and deployed PrepLens, a full-stack placement-prep platform (React, Node/Express, MongoDB, Vercel + Render) used by ~2,000 students, saving each an estimated 10+ hours of interview preparation. Owned it end-to-end, from database design and REST API to UI, search and a feedback-driven redesign.
  • Designed a pre-publication moderation workflow with an admin review queue: every submission is approved, or rejected with feedback to the author, and each decision is audited by reviewer. Added company-name deduplication (alias matching plus admin merge/reject) to keep search filters clean.
  • Implemented role-based access with separate student and admin sign-in over Google OAuth, restricted to the college domain plus an admin allowlist. Server-side hashed sessions and API-level anonymity protect author identity, backed by 117 automated tests covering auth, moderation and data integrity.
ReactViteTailwind CSSNode.jsExpressMongoDBGoogle OAuthREST APIVercelRender

ForestLens – Tree Crown Detection & Canopy Area Estimation

  • Built and deployed a geospatial ML web application for individual tree crown detection and canopy area estimation from high-resolution satellite imagery using DeepForest, Rasterio, and Streamlit.
  • Diagnosed and fixed a hidden DeepForest tile-scale dependency that fragmented crowns across tile sizes, improving crown geometry consistency and pipeline reproducibility.
  • Investigated native-resolution detection failures through documented scale matching, reaching 71 trees/ha with 5.2 m crowns while explicitly reporting uncertainty and withholding unsupported measurements.
PythonPyTorchDeepForestRasterioStreamlitComputer VisionGeospatial ML

Optiforge Neural Options Pricing

  • Built a neural options pricing stack with LSTM models benchmarked side-by-side against Black–Scholes for fair quantitative comparison.
  • Layered in GARCH-style volatility and analytical pricing so classical and learned estimators can be evaluated on the same surfaces.
  • Packaged a Streamlit demo for interactive exploration and faster iteration on model behavior and error profiles.
PythonLSTMGARCHFinancial ModelingDeep Learning

AI-Powered File Organizer with OS-Level System Calls

  • Connected low-level C system calls to a Next.js/Node full-stack app for file operations, natural-language commands, and cross-format semantic search (PDF, images, text).
  • Implemented AI-assisted auto-tagging and a multi-step autonomous organizer with explicit safety rules and user confirmation for risky actions.
  • Tuned the agent for predictable tool use and context limits so long-running organization tasks stay controllable in real directories.
COperating SystemsNext.jsNode.jsAISemantic SearchLLMFull StackSystem Design
05Academics

Education

  • Bachelor of Technology (Artificial Intelligence)

    Newton School of Technology, Rishihood University

    Delhi NCR, India

    2023 – 2027

    7.05 / 10.0

  • Intermediate (Class XII)

    Malviya Convent School

    Jaipur, Rajasthan

    2021 – 2022

    83.0%

  • Matriculation (Class X)

    St. Edmund's School

    Jaipur, Rajasthan

    2019 – 2020

    91.1%

06Recognition

Honors & Awards

  • JEE Advanced '23 — AIR 6522

    Jun 2023

    Issued by IIT Delhi

    Secured All India Rank (AIR) 6522 in JEE (Advanced) 2023 among more than 185,000 candidates. Score: 142/360. OBC-NCL category rank: 1390.

  • JEE Mains '23 — AIR 24875

    Apr 2023

    Issued by NTA (National Testing Agency)

    Secured All India Rank (AIR) 24875 in JEE Main 2023 with a 97.84 percentile. OBC-NCL category rank: 6397.

07Toolkit

Skills

ML & AI

Machine LearningDeep LearningNLPComputer VisionTime Series ForecastingScikit-learnTensorFlowPyTorch

Generative AI

LLMsRAGLangChainLangGraphOpenAI API

Languages & Scientific Computing

PythonSQLJavaScriptTypeScriptJavaOOPNumPyPandasMatplotlib

Databases

PostgreSQLMySQLMongoDBSupabaseFirebasePrisma ORM

Web, Backend & Tools

ReactNext.jsNode.jsExpressREST APIsSystem DesignGit/GitHubTailwind CSSHTMLCSSSwaggerUI/UX

Core CS

Data Structures & AlgorithmsProblem-solving

Soft skills

ResearchTeamworkCommunication skills
08Learning

Certifications

  • Ethical Hacking · CourseraLink

    May 2025

    Completed hands-on training in ethical hacking, focusing on identifying vulnerabilities, penetration testing, and securing systems against cyber threats.

  • DSA Course · Apna CollegeLink

    October 2023

    Completed a hands-on DSA course, enhancing problem-solving skills, code optimization, and foundational knowledge in algorithms.

09Activity

Shipping & solving, live

GitHub contributions

@RAVIYADAV6522
GitHub contribution chart for RAVIYADAV6522
10Beyond code

Achievements & Activities

  • IEEE PES Energy Shark Tank 2026 (18th APPEEC, Singapore): FlexGrid pitch selected among the Top 5 in the IEEE PES YP Industry Innovation Session; awarded a Certificate of Achievement for the most outstanding innovative solution and placed 3rd overall.
  • 1st Place, NST Startup Foundry 2026: Won with Jarvis, an AI-powered personal intelligence system, pitched to Google Cloud and Microsoft for Startups.
  • Mentorship: Mentored 10+ students in Data Structures & Algorithms.
  • IEEE Student Member (2026): Student Member in good standing.
  • Jaipur Under-16 Cricket Team: Represented Jaipur as an opening batsman and part-time wicketkeeper.
  • More Than Me: Volunteer supporting children in need.
11Contact

Get In Touch

I'm currently open to new opportunities. Whether you have a question, a project idea, or just want to say hi — my inbox is always open!

© 2026 Ravi Yadav · Built with Next.js, Tailwind & Framer Motion

Loading portfolio…