Divyansh Yadav | Software Engineer & CSE Undergrad at IIT Jodhpur

A Software Developer pursuing his B.Tech in Computer Science and Engineering at IIT Jodhpur.

I like to try new and different things in my life, no matter how random or bizarre they may seem to the general eye.

Divyansh Yadav

Hey! I am a Tech Enthusiast with a keen interest in Software Development, Cybersecurity and System Design. I like to design and build secure and scalable software systems that do not break under pressure. I have recently grown very fond of the ideas of Distributed Systems and Operating Systems, and I am currently adding them to my knowledge arsenal.

Outside the realm of computer sorcery, I enjoy doing ice-skating and playing basketball. I enjoy watching movies, web series and animes in the time when I think I should probably be sleeping. I also love to participate in debates and improvs which help me bring the necessary drama and chaos in my life.

2024 - Present

BTech in Computer Science

Indian Institute of Technology, Jodhpur

  • Maintaining a strong academic record with a current CGPA of 8.55.

  • Mentored 12 freshers throughout their first year as a Student Guide for the Wellbeing Committee.

  • Managed digital infrastructure as Tech Coordinator for Alumni Affairs.

  • Directed flagship technical competitions as Assistant Head for Prometeo 2026.

  • Represented the institute at Inter IIT Tech 14.0.

2022 - 2024

Class XII (CBSE)

Paramount International School

  • Graduated with 95% in the CBSE Class XII Board Exams.

  • Secured overall Rank 3 in the school.

2010 - 2022

Class X (CBSE)

Modern Convent School

  • Achieved 96.2% in the CBSE Class X Board Exams.

  • Honored with the highly coveted Student All-Rounder Award.

  • Led athletic teams and sporting events as the Vice Captain for Sports.

September 2025 - November 2025

Software Development Intern

Tech Swarm

  • Engineered a reliable drone-to-ground telemetry communication system utilizing MQTT and Protocol Buffers to ensure seamless data transmission.

  • Architected a high-throughput asynchronous Kafka ingestion pipeline, integrating InfluxDB for downsampled storage to optimize and process geospatial queries.

November 2025 - December 2025

Computer Networks and Cryptography Lead

Inter IIT Tech 14.0, Patna (Qtrino Labs)

  • Implemented a secure DTLS 1.3 communication protocol tailored for bare-metal RISC-V IoT devices, achieving zero dynamic memory allocation to prevent memory fragmentation and ensure deterministic performance in a constrained environment.

  • Designed an asynchronous cryptographic data pipeline integrating state-of-the-art Post-Quantum Cryptography (CRYSTAL-Kyber and CRYSTAL-Dilithium) and pre-shared key (PSK) authentication, utilizing lock-free ring buffers to optimize high-speed packet ingestion, compress the memory footprint, and sustain low-latency secure transmissions.

September 2025 - January 2026

Software Development Intern

MyGyan

  • Engineered an intelligent, LLM-driven multimodal pipeline capable of dynamically generating personalized video lectures and quizzes matched to custom learning goals, implementing advanced semantic caching to efficiently store and reuse assets.

  • Developed a custom LLM routing strategy leveraging RouteLLM to intelligently direct prompts based on domain proficiency, cost constraints, and request urgency to achieve optimal accuracy and latency.

  • Architected and orchestrated scalable, containerized microservices using Amazon EKS, and designed event-driven serverless routines on AWS Lambda to ensure highly cost-effective, high-availability infrastructure.

  • Automated secure CI/CD pipelines using GitHub Actions and built centralized observability dashboards with AWS CloudWatch to streamline deployments and accelerate incident response times.

OmniView

OmniView is an on-device screen intelligence system acting as a secure photographic memory. Operating as a background service, it captures screen frames while discarding duplicate captures. The screenshots are then processed to extract context, which is further vectorized and stored locally. Users can query past screen history via an on device LLM.

AndroidKotlinMediaProjection APIpHashAccessibility APIOCRMobileBERTGemma 4ObjectBox

AgentLens

AgentLens is an autonomous brand audit engine designed for Generative Engine Optimization (GEO) to evaluate how effectively websites are discovered, interpreted, and cited by AI agents and answer engines. Built to the agentskills.io standard for the Adobe University Hackathon, it analyzes crawler accessibility, structured data grounding, and content extraction to uncover AI visibility gaps. The system generates prioritized audit reports with actionable remediations to maximize brand presence in LLM responses.

PythonAgent SkillsJSON-LD

MiniTrue

MiniTrue is a decentralized, leaderless time-series database designed for high-throughput IoT telemetry workloads. Operating as a symmetric multi-node cluster, it gossips over TCP and uses a consistent hash ring for deterministic data placement. Ingested data is cached in-memory with pre-aggregated metadata, then compressed and flushed to a custom columnar disk format for fast local and distributed query execution.

GoGossip ProtocolConsistent HashingMerkle TreesGorilla CompressionParquet Storage

Seiton

Seiton is an autonomous 3D cargo packing optimizer that calculates optimal placements for heterogeneous cargo boxes inside a container, maximizing volume utilization. It uses OpenCV to extract physical dimensions from raw 2D images and routes them through a high-performance C++ heuristic optimization engine. The system also features a real-time digital twin visualization showing box orientations and placement order in a WebGL-powered environment.

ReactThree.jsGoC++PythonOpenCVGSAP

BitVision

BitVision is designed to predict Bitcoin price trends using raw OHLCV market data and technical indicators like RSI and Bollinger Bands. After training and benchmark testing multiple predictive models—including SVMs, Naive Bayes, Decision Trees, LSTMs, XGBoost, and Neural Networks—Ridge Regression was selected as the final model due to yielding the lowest Mean Absolute Error and highest R-squared score on out-of-sample data.

PythonScikit-LearnRidge RegressionPandasNumPyStreamlit

ClarifAI

ClarifAI is an AI-powered learning platform that allows users to upload their coursework and study notes. To help students prepare for exams, it implements a Retrieval-Augmented Generation pipeline using sentence-transformers and Pinecone to ground answers directly in the uploaded documents. The system also generates custom assessment papers and personalized mock tests to track learning progress.

DjangoFlutterPineconeGoogle GeminiSBERTSQLite

The Hive

The Hive is a multi-device digital wellness and productivity platform that tracks browsing and application habits across Chrome and Android. It uses on-server transformer models to generate semantic embeddings of domains and apps, classifying them automatically via pgvector and HNSW similarity searches. Daily activities are processed through a Bayesian rating system to score productivity trends and gamify habit building.

KotlinNode.jsPostgreSQLpgvectorChrome ExtensionTransformers

Destyn

Destyn is a college community application designed to connect local students sharing similar interests across nearby campuses. The platform enables users to discover common interests and helps them receive personalized activity recommendations for hanging out in real life. Additionally, it features an anonymous community gossip page and facilitates real-time chat and interactions.

Next.jsNode.jsMongoDbCloudinaryWebSocketsGoogle Gemini

Alumni Yearbook Portal

Alumni Yearbook Portal is a portal designed for the graduating batch of IIT Jodhpur. The platform enables students to upload photo memories and write messages to peers, compiling them automatically into personalized digital yearbooks for each graduate. To drive engagement, it features batch-wide polling alongside private messaging that allows user messages and quotes to be directly integrated into yearbooks.

ReactPostgreSQLAWS S3Node.js

PathViz

PathViz is a desktop application designed to visualize maze generation and pathfinding algorithms. Built in C, the system generates random grid mazes and solves them in real-time using Dijkstra's and A* search algorithms. The graphical user interface is built natively using GTK, allowing users to watch the step-by-step traversal and compare algorithm efficiencies visually.

CGTKDijkstraA*

Loading thoughts & insights...

Languages

C
C++
Python
JavaScript
TypeScript
Go
Kotlin
Assembly

Development

HTML
CSS
Tailwind
React
Next.js
Node.js
Express
Django

Databases

MySQL
PostgreSQL
MongoDB
Redis

DevOps & Infrastructure

Git
GitHub
Docker
Kubernetes
Kafka
RabbitMQ
GitHub Actions
Prometheus
Grafana
Terraform

Cyber Security & Networking

Postman
Wireshark
Burp Suite
Metasploit
OWASP
Splunk

Check out my presence on other platforms and please do not hesitate to follow me xD :)

Feel free to reach out if you have a question, or just want to connect.