Hey, I'm Anudeep.
I bridge the gap between machine learning pipelines and production-grade full-stack applications. Leveraging AI-assisted workflows to ship high-impact software, from adaptive traffic control systems to accessibility-first transit platforms.
Philosophy
AI-Driven. Full-Stack.
Product Focused.
Currently pursuing a B.Tech in Computer Science at VNR VJIET (CGPA: 9.37) and working as a Full Stack Developer Intern at PSNM Innovations.
What problems do I enjoy solving?
I am drawn to complexity where machine learning meets physical and digital infrastructure. Whether it is engineering graph traversal constraint models for an 8,737-station railway network, training computer vision pipelines to analyze live traffic congestion, or building AI-assisted learning interfaces, I focus on systems that optimize resources and improve accessibility.
How do I build?
With extreme velocity, powered by modern AI-assisted engineering workflows. I leverage tools like Cursor, Claude Code, and GitHub Copilot to handle repetitive development, refactoring, and multi-file changes. This shifts my focus to what matters: high-level architecture planning, rigorous constraint optimization, and high-fidelity user experiences.
What motivates me?
Measurable, real-world utility. I believe developer-builders should own products end-to-end, from custom ML models to production-grade deployments. Seeing an adaptive traffic signal engine cut vehicle wait times by 29.89% or watching a civic reporting tool help fifty active users resolve local infrastructure issues is what drives me.
Case Studies
Building systems that optimize infrastructure.
AI-Based Urban Traffic Flow Optimization System
The Challenge
Conventional fixed-timing traffic signals lead to severe congestion and empty-intersection waiting times because they fail to adapt to real-time changes in vehicles and queue volumes.
The Solution
Engineered an adaptive traffic control system using YOLOv8 and OpenCV for real-time queue detection. Integrated a feedback control loop adjusting green-light durations (+10s to -11s per direction) based on a blend of 70% predictive historical congestion trends and 30% live queue density metrics.
Measurable Impact
- •Reduced total vehicle waiting times by 29.89% compared to traditional fixed schedules.
- •Validated system functionality through highly realistic SUMO (Simulation of Urban MObility) runs.
- •Awarded First Runner-Up at VNR Designathon 2026 out of 2,800+ total registrations.
SUMO Adaptive Traffic Simulation
Active Optimization Logic: YOLOv8 Congestion Pipeline
YOLOv8 congestion pipeline is dynamically adjusting light durations between +10s and -11s depending on relative queue weights.
Civix — Civic Issue Reporting Platform
The Challenge
Municipal infrastructure reporting often suffers from sluggish verification times, duplicate submissions, and inaccurate location tags, making resolving local community hazards inefficient.
The Solution
Built a location-aware full-stack dashboard utilizing React and Next.js. Deployed an automated AI-validation pipeline validating coordinates, checking for duplicate tickets, and filtering reports via location tagging and object detection.
Measurable Impact
- •Won 1st Place at Webathon 4.0 Hackathon against 1,000+ competitors.
- •Successfully tracked issues and verified coordinates for 50+ active testers.
- •Significantly reduced reporting overhead using automated verification pipelines.
Civix AI Validation Pipeline
Webathon 4.0 Gold Medal Platform Demo
Submit New Issue
Live Issues Feed
AI Railway Optimization Platform
The Challenge
Routing algorithms usually focus purely on travel time, ignoring the accessibility needs of disabled travelers, which isolates passengers from key parts of public transit systems.
The Solution
Engineered an accessibility-first routing engine mapped onto an 8,737-station network. Formulated paths using Time-Expanded Graphs and resolved station-facility constraints using the Google OR-Tools CP-SAT solver.
Measurable Impact
- •Ensured disability-friendly journey mapping for 64% of the Indian railway network.
- •Structured paths that guarantee wheelchair ramps, audio guides, or lift assistance at every node.
- •Substantially reduced routing calculation times across massive datasets.
CP-SAT Access Router
Solver parameters: 8,737 stations database
Route verified using CP-SAT solver. Checked disability accommodations across network paths.
Employment
Professional Experience
Full Stack Developer Intern
Traditional study resources (textbooks, notes, handouts) are highly passive, dense, and non-interactive, which limits student retention and lacks personalized study guidance.
Architecting and shipping StudyAI, an AI-powered learning engine that parses study materials and dynamically transforms them into explanations, quizzes, interactive flashcards, and adaptive study calendars. Engineered analytics dashboards and moderation systems following rigorous SDLC frameworks.
Significantly accelerated product shipping times by integrating AI-driven code refactoring, system architecture planning, and debugging pipelines while maintaining strict branching and PR reviews.
Product Team Member
Local municipal crews lacked accurate, real-time spatial data on road hazards, while citizens lacked a fast, frictionless tool to report street damage programmatically.
Co-developed and tested a community-focused location-aware reporting interface built for pothole detection, GPS logging, and hazard severity mapping.
Ran extensive validation testing and customer feedback iterations, directly translating usability gaps into concrete UI updates to improve user acquisition.
Capabilities
Technical Skill Matrix
AI-Assisted Development
Advanced engineering flows utilizing Large Language Models to accelerate code iteration and system planning.
Technologies & Practices
Milestones
Achievements & Certifications
1st Place — Webathon 4.0 Hackathon
AI-enabled civic-tech reporting infrastructure. Placed 1st among 1,000+ total active participants.
First Runner-Up — VNR Designathon 2026
AI-Based Urban Traffic Flow Optimization System. Ranked in the top 30 submissions out of 2,800+ registrations.
2-Star CodeChef Coder
1420 RatingAchieved an official peak rating of 1420 on the global CodeChef coder scoreboard.
Finalist — Solution Sprint Ideathon
Advanced to final validation rounds with an optimization-first software prototype.
Professional Certifications
Community & Leadership
Action Committee Member, Turing Hut (Coding Club)
Selected through a competitive programming contest and technical interview process from 900+ applicants. Conducted coding contests and tech workshops.
Member, Computer Society of India (CSI)
Supported developer community programs and participated in technical hackathons.