// CS&E (AIML) Undergraduate · SRMIST

Phani
Koushik

Builder of Intelligent Systems.

I build AI-powered, full-stack products — from CNN waste classifiers to multilingual voice assistants. Currently pursuing B.Tech at SRMIST with a CGPA of 9.11.

View Projects →Get In Touch

Available for internships · Class of 2027

SRMIST '27

AI/ML · B.Tech '27

AI/MLCNNReactPython

Education

Aug 2023 – Present

SRM Institute of Science and Technology (SRMIST), Kattankulathur

B.Tech in Computer Science Engineering — Specialization in AI/ML

CGPA: 9.11 / 10

Jun 2021 – Apr 2023

Gitam Junior College, Tirupati

Intermediate — Mathematics, Physics, Chemistry (MPC)

Grade: 80%

About

I'm Phani Koushik Reddipalli, a final-year B.Tech Computer Science Engineering undergraduate specializing in AI / ML at SRMIST, Kattankulathur, seeking AI/ML Engineer roles.

I've built a hybrid-retrieval RAG system (FastAPI, pgvector, BM25, reranking), a multimodal voice-first Generative AI app on Gemini and Supabase, and a Random Forest/SVM classifier — with cloud deployment on AWS, Docker, and Vercel.

When I'm not coding, you'll find me sharpening problem-solving skills through DSA challenges or exploring new frameworks. I also hold an NCC 'A' Certificate and am an NMMS scholar.

EnglishTeluguKannadaHindiGermanTamil

9.11

CGPA @ SRMIST

AI/ML Projects

Certifications

Languages Spoken

Skills

// Languages

PythonJavaC++CJavaScriptTypeScriptSQL

// AI / ML

Machine LearningDeep LearningFeature EngineeringModel EvaluationRandom ForestSVMCNNRNN / LSTMPyTorchTensorFlowScikit-learnPandasNumPy

// GenAI / LLMs / NLP

LLM APIs (Gemini)Multimodal AIPrompt EngineeringStructured OutputsRAGEmbeddingspgvectorBM25Hybrid SearchRerankingRAG EvaluationAI AgentsMultilingual NLP

// Frontend

React.jsViteHTML5CSS3Tailwind CSSshadcn/uiWeb Speech API

// Backend / APIs

FastAPINode.jsREST APIsSupabase Edge Functions

// Databases

PostgreSQLpgvectorMySQLSupabase

// Cloud / DevOps

AWS (EC2, S3, IAM)DockerCI/CDVercel

// Tools

GitGitHubPostmanMATLAB

Projects

// 001

EcoVoice AI

Environmental Intelligence System

Full-stack, voice-first Generative AI app integrating AI waste analysis, water-conservation guidance, and location-aware sustainability recommendations.

Multimodal waste-image analysis with Gemini 2.5 Flash via Supabase Edge Functions
Structured 6-field output: classification, material, contamination risk, confidence, disposal & eco-tips
Flow: text/voice/image → React → Edge Function → Gemini → on-screen + spoken response
Multilingual voice assistant in 5 languages with streaming & custom speech queue

ReactTypeScriptViteTailwind CSSSupabasePostgreSQLGeminiWeb Speech API

Live Demo ↗Source ↗

Languages

// 002

RAGForge

Intelligent Technical Knowledge & Debugging Assistant

End-to-end Retrieval-Augmented Generation assistant for technical Q&A and debugging, grounded in GitHub repositories and technical documents.

Structure-aware chunking → embeddings → hybrid retrieval (pgvector + BM25) → reranking → Gemini
Citation-grounded Q&A, retrieval inspection, and debugging mode
Evaluated on Recall@K, MRR, faithfulness, answer relevance, and latency

ReactTypeScriptPythonFastAPISupabasePostgreSQLpgvectorGemini

RAG

Hybrid

// 003

Health Risk Prediction

Supervised Machine Learning Pipeline

Classification pipeline from raw health data to risk prediction, comparing Random Forest and SVM on real-world data.

Data cleaning → feature selection → Random Forest & SVM → evaluation
~80% accuracy; analyzed why accuracy alone misleads on imbalanced data
Model chosen on accuracy, precision, and recall; reproducible Pandas/NumPy workflow

PythonScikit-learnPandasNumPy

80%

Accuracy