May 11, 20261 min read
Test
BackendUsed with:Node.jsFrameworks like:Express.jsNestJSExample uses:APIsServersAuthentication systemsDatabases
fdfdsfsd
Hi, my name is
Jeevith Gangadhar.
I'm a |
A passionate software engineer with research publications in multilingual phishing detection and affordable voice AI systems. Currently pursuing B.Sc in Computer Science at MAHE while building production software at Colcord.
Latest Research
UniPhish: Unicode-Native 1D-CNN for Multilingual Phishing URL Detection
98.7% accuracy across 6 script families | 487,523-URL multilingual benchmark
2
Research Papers
98.7%
UniPhish Accuracy
5+
Hackathon Wins
85%
Cost Reduction
01.About Me
Hello! I'm Jeevith, a Full Stack Developer and AI/ML Researcherbased in Bengaluru, India. I'm passionate about building software that solves real-world problems, particularly in the domains of cybersecurity, multilingual AI, and healthcare technology.
My journey in tech started with building a University Companion App that was later acquired by Colcord. This experience taught me end-to-end product development, from ideation to scale. Today, I lead technical initiatives at Colcord while pursuing cutting-edge research in AI.
At Kasturba Hospital, I work on AI-driven healthcare projects, building ML pipelines that analyze Electronic Health Records for clinical decision support. My research has led to publications on UniPhish (multilingual phishing detection with 98.7% accuracy) and affordable multilingual voice agents for Indic languages.
I believe in making technology accessible. My voice AI research achieved an 85-90% cost reduction over commercial solutions, enabling deployment in resource-constrained environments across India.
Technologies I work with:
Frontend
- ▹React
- ▹Next.js
- ▹React Native
- ▹Tailwind CSS
- ▹TypeScript
Backend
- ▹Node.js
- ▹Express.js
- ▹Flask
- ▹Supabase
- ▹REST APIs
AI/ML
- ▹PyTorch
- ▹TensorFlow
- ▹RAG Workflows
- ▹LLM Orchestration
- ▹Whisper ASR
Databases
- ▹PostgreSQL
- ▹Firebase
- ▹Redis
- ▹FAISS
Education
B.Sc Computer Science Engineering
ICAS Manipal, MAHE (2024 - Present)
GPA: 3.0/4.0 (Till 3rd Semester)
High School
Baldwin Boys High School
Grade: 93% | A+ in Computer Science
Key Coursework
Operating SystemsData StructuresDesign & Analysis of AlgorithmsMachine LearningPrinciples of CryptographyComputer Networks
02.Research Publications
My research focuses on making AI accessible and secure for underserved populations, from multilingual cybersecurity to affordable voice technology for Indic languages.
IEEE (Under Review) · 2025
UniPhish: Unicode-Native 1D-CNN for Multilingual Phishing URL Detection in Messaging Applications
Jeevith Gangadhar, Raunak Khandelwal, K. Shashank Reddy, Sudhina Kumar G. K., Hareesha K. S.
Phishing attacks distributed through instant messaging services represent a growing cyber threat for billions of non-English speaking users. This paper presents UniPhish, a Unicode-native parallel multi-kernel 1D-Convolutional Neural Network that encodes every URL as a sequence of UTF-8 bytes, achieving complete Unicode coverage through a vocabulary of only 256 tokens.
Key Results
- ▹98.7% macro-average accuracy across 6 major script families
- ▹98.9% macro-F1 score with 0.43ms prediction latency on CPU
- ▹668× faster than Multilingual BERT at equal accuracy
- ▹96.1% accuracy against Unicode homoglyph attacks
- ▹4.2 MB compressed model for sub-1ms mobile inference
Dataset
MultiPhishDB: 487,523-URL multilingual benchmark (κ = 0.91)
Technologies
1D-CNNUTF-8 EncodingPyTorchTensorFlow LiteAdversarial TrainingKnowledge Distillation
Manipal Institute of Technology · 2025
Affordable Multilingual Voice Agents: A Self-Hosted Architecture for Indic Languages
Jeevith Gangadhar, Dr. Nagaraj Naik
This paper proposes a voice agent platform for multilingual Indian users, achieving 85-90% cost reduction compared to commercial alternatives while maintaining conversation quality. The system supports Hindi, Tamil, Telugu, Kannada, Tulu, and English with automatic language identification.
Key Results
- ▹85-90% cost reduction (₹0.95/min vs ₹7-15/min commercial)
- ▹742ms end-to-end latency, 488ms time-to-first-sound
- ▹Supports 6 Indic languages natively
- ▹Integrated emotional distress detection for safety-critical scenarios
- ▹37% WER reduction for Tulu minority language
Architecture
MMS-LID → Whisper ASR → LLaMA 3.2 LLM → Piper TTS
Technologies
WhisperLLaMA 3.2Piper TTSWebRTCTwilioMimi Neural CodecRedisFAISS
3.4B
Users Protected
Non-English speakers
487K
URLs Analyzed
MultiPhishDB dataset
6
Script Families
Unicode coverage
6
Indic Languages
Voice AI support
03.Where I've Worked
1 / 3
Software Developer Intern @ Ericsson
May 2026 – September 2026Bengaluru, IndiaStarting Soon
180+
Countries Served
5G / Cloud
Focus Area
DevOps
Role
About Ericsson
Working with Ericsson, a Swedish-based multinational corporation (MNC) specializing in telecommunications, networking, cloud, and 5G technologies, serving clients across more than 180 countries.
My Role
Working in a DevOps role with exposure to Linux, CI/CD pipelines, automation, cloud infrastructure, monitoring tools, and enterprise-level deployment workflows in a global technology environment.
Key Contributions
- ▹Building and maintaining CI/CD pipelines for enterprise-grade telecommunications infrastructure deployments across global environments
- ▹Working with Linux systems and automation tools to streamline deployment workflows and reduce manual intervention in release processes
- ▹Contributing to cloud infrastructure management using Kubernetes, Docker, and Terraform for scalable 5G network solutions
- ▹Implementing monitoring and observability solutions using Prometheus and Grafana for mission-critical telecom systems
- ▹Collaborating with international teams across multiple time zones on cutting-edge 5G and networking technologies
- ▹Gaining exposure to enterprise-level security practices and compliance requirements in telecommunications infrastructure
Technologies Used
LinuxCI/CDKubernetesDockerTerraformJenkinsAnsiblePrometheusGrafanaGitPythonBash
scroll to explore all companies ↓
04.Skills & Technologies
Practical engineering strengths shaped by production shipping, hackathon speed, research implementation, and multi-repo experience across AI, full-stack, and voice systems.
Frontend
React / Next.js95%
TypeScript90%
Tailwind CSS92%
React Native88%
HTML5 / CSS395%
Backend
Node.js / Express88%
Flask / Python85%
Supabase90%
REST APIs92%
WebRTC / Twilio80%
AI / ML
PyTorch / TensorFlow85%
LLM Orchestration88%
RAG Workflows85%
Whisper / Voice AI82%
1D-CNN / NLP80%
Databases
PostgreSQL88%
Firebase / Firestore90%
Redis78%
FAISS (Vector)75%
MySQL82%
Languages & Specialized Tools
Languages
PythonJavaScriptTypeScriptC++JavaSQLDart
Voice AI
Whisper ASRPiper TTSOllamaMMS-LIDMimi Codec
Security
Phishing DetectionUTF-8/UnicodeAdversarial TrainingRLS
DevOps
GitDockerGitHub ActionsCI/CDTFLiteVercel
Other things I'm good at:
Product ThinkingSystem DesignTechnical WritingResearch & AnalysisTeam LeadershipMentoringAPI DesignPerformance OptimizationSecurity Best PracticesAgile / Scrum
04.Projects
Total Repos
101
Public
41
Private
60
Top Language
TypeScript
Featured Projects
ICAS Student Platform
Enigma Hackathon 2025 - Team AXION
Comprehensive web application revolutionizing the ICAS 2+2 transfer program experience. Features intelligent automation for foreign transfer processes, structured peer/alumni connections, and streamlined event management.
Full Stack Lead
TypeScriptReactNext.jsExpress.js
University Companion App (YOGO)
Acquired by Colcord
Cross-platform student companion app simplifying daily campus workflows including food planning, ride sharing, marketplace, bill splitting with OCR, timetable management using Gemini AI.
50% reduction in coordination
React NativeExpoFirebaseGemini AI
Showing 15 of 101 projects
SHAP ML Pipeline
End-to-end hand-injury ML workflow with SHAP feature importance, XGBoost, Random Forest for MHISS/ATTHA/MUES scores.
PythonXGBoostSHAP
Voice AI Model
Multi-tier voice AI for Indian languages with Whisper ASR, LLaMA 3.2, Piper TTS. 742ms latency.
PythonPyTorchFlask
Speech-to-Speech Transformer
Unified de-novo S2S transformer for Indian languages with under 300ms latency.
PythonPyTorchSNAC
LiveLens RAG News
Real-time adaptive news intelligence using Pathway streaming RAG.
DataQuest 2026 | IIT Kharagpur
PythonPathwayNode.js
NASA Asteroid Impact API
ML system for asteroid impact prediction using NASA NEO data and Monte Carlo.
NASA Space Apps
PythonFastAPIML
Demo THIT Voice AI
Real-time Voice AI with Faster-Whisper STT, Llama LLM, Edge TTS.
PythonFastAPIWhisper
Model Arena
Competitive AI model evaluation platform with multiple game modes.
Vercel AI Hackathon
TypeScriptReactVercel AI
05.Achievements & Recognition
Hackathons, certifications, and academic excellence
Hackathons & Competitions
IIIT Hyderabad Hackathon
Finalist
Voice AI Agent for Apollo Hospitals
Built using speech-to-text, NLP, and LLM-based reasoning to assist patients with queries and appointment workflows.
MIT Manipal Hackathon
4th Place
iWitness - Anonymous Crime Reporting
Platform focused on user trust, privacy, and accessible UX design for secure crime reporting.
IIT Roorkee Frontend Hackathon
Finalist
Sponsor-focused Event Website
Built responsive UI with strong performance and clear brand communication for event sponsors.
NASA Space Apps Challenge
Participant
Data Visualization & Simulation
Leveraged real NASA datasets for scientific storytelling and space exploration education.
Certifications
Machine Learning Specialization
Andrew Ng, DeepLearning.AI
Coursera
Supervised LearningNeural NetworksDecision TreesRecommender Systems
Generative AI with Large Language Models
DeepLearning.AI & AWS
Coursera
LLM TrainingFine-tuningRLHFPrompt Engineering
Academic Excellence
A+ in Computer Science
Board Examinations
93% in High School
Baldwin Boys High School
GPA 3.0/4.0
ICAS Manipal (Till 3rd Sem)
Currently Exploring
System DesignAdvanced TypeScriptCloud ArchitectureMLOpsKubernetes
06.Latest from the Blog
Thoughts on software development, AI, and building products