Chaitanya

Chaitanya

AI/ML Engineer


About

I am developer making things to make people life easier. I love to expand my horizons and learn more about new technologies. I am always open to new ideas and always be a part of things that intrigue my interest

Skills
JavaScript
TypeScript
Next.js
Express
Tailwind CSS
C
REST API
Vitest
Playwright
Docker
OAuth
AI Agent
LLM API Integration
prompt design
multi-agent system
Experience

Projects

AnimeMangaClub

A full-stack website for an anime and manga club: members give titles a verdict, keep watch and reading lists, like each other's reviews, get personal suggestions that explain themselves, and see what the club leads pick each week. Verdicts, not numbers. Members rate titles Skip, Timepass, Go for it or Perfection, each marked by its own colour. Each title shows the club's verdict with a semicircle gauge of how the votes split. Suggestions that say why. "For you" blends a taste profile (genres, weighted themes and preferred formats), "members with similar taste loved this", AniList and club quality, one pick per franchise, and a reason on every card ("More Attack on Titan, which you loved", "Because you like Revenge and Military stories"). Bring your list. Import an AniList anime and manga list in one go; statuses and your scores feed your suggestions straight away. Search that speaks Japanese titles. "shingeki no kyoujin", "kimetsu no yaiba" or "SnK" find the English-titled entries, forgiving spelling and missing words. Club leads' weekly picks, planned ahead privately and published when the week starts. Accounts done properly: email and password or Google, email confirmation, password reset, breached-password checks, a "new sign-in" email with a way to lock an intruder out, and no way to probe who is a member. Accessible and fast: WCAG 2.2 AA checked by automated axe scans on every page; Lighthouse 100 for accessibility, best practices and SEO; 99–100 performance on desktop.

TypeScript
Next.js 16
React 19
Hono
PostgreSQL (Neon)
Drizzle ORM
Better Auth
Tailwind CSS
GSAP
three.js
Playwright
Docker

AirMouse

Control your Windows mouse with bare-hand gestures in front of a webcam. AirMouse tracks the 21 joints of your hand, draws them over the camera view, and turns gestures into cursor movement, clicks, drags and scrolling. It runs quietly in the system tray.The app window shows the live camera with your hand drawn on it, a guide that lights up the gesture you're making, and settings. Fit area to my reach measures how far your hand comfortably moves and sizes the control area to match. Hotkeys (work from any app): Ctrl+Alt+M turns hand control on or off, Ctrl+Alt+Q quits. Emergency stop: move the physical mouse into any screen corner. Everything runs on your computer. Camera images are never saved or sent anywhere. The window talks to the tracking engine over a server bound to 127.0.0.1 that requires a random per-session token, so web pages can't connect to it. Settings are stored in %APPDATA%\AirMouse.

Python

FeedBack Analyser

Give it a business name and a location. It scrapes that business's public reviews, throws away the junk, groups what's left into themes, and scores each theme from −5 to +5 with a reason and a concrete next step — so an owner sees what to fix first and what not to break, not just a star rating. It also remembers. Every run is stored, so the next one can say "crowding was flagged 10 times last month, 5 this month" instead of restating one batch.

Python
JavaScript
CSS
HTML
Dockerfile

Mail Sentry

AI-powered email threat detection, geolocation & forensic intelligence platform. Connects to a user's Gmail (Google OAuth, read-only), analyzes every message for phishing, sender spoofing / impersonation, malicious URLs & attachments, social engineering and Business Email Compromise, assigns a 0–100 fraud/risk score with a human-readable explanation, and adds geolocation + forensic intelligence per mail (originating IP, mail-server hop timeline, SPF/DKIM/DMARC authenticity).

TypeScript

EUROSIM

EUROSIM is a real-time geopolitical simulation where six AI-controlled European nations interact autonomously on an interactive SVG map. Each nation has a distinct personality, maintains trust scores toward every other nation, remembers past interactions, and uses a Large Language Model (Meta LLaMA 3.1 8B via Featherless AI) to make strategic decisions. You can trigger events (attacks, trade deals, alliances, sanctions) and watch the AI nations react — or start the autonomous simulation and observe emergent geopolitical narratives unfold in real time. Key Features Real AI Decisions — Each nation's response is generated by LLaMA 3.1 8B with full context (trust, alliances, memory, personality) Transparent AI — Every decision shows its source (AI / Rule-based / Autonomous) with full reasoning visible in the Strategic Analysis panel 6 Nations with Distinct Personalities — France (diplomatic), Germany (diplomatic), UK (opportunistic), Russia (aggressive), Poland (defensive), Italy (opportunistic) Trust + Alliance + Memory Systems — Three interlocking systems create emergent behavior: trust shifts influence alliances, alliances affect decisions, memories create grudges and loyalties Real-World Event Injection — Bright Data integration pulls real geopolitical news to influence the simulation Military Command Center UI — Dark-themed interface with animated event arrows, nation dossiers, and live debug overlay Graceful AI Fallback — If the AI API fails, a comprehensive rule-based system takes over seamlessly

JavaScript
CSS
HTML
Education

Btech

GITAM Hyderabad

Aug 2025 to Apr 2029
Certifications

B Sri Krishna Chaitanya

Forge Inspira

Mar 2026
Chaitanya · DevFest Hyderabad 2026