Sai Ruthvik hawkh

`> whoami`

class SaiRuthvik:

    name        = "Sai Ruthvik"
    alias       = "hawkh"
    location    = "IIT Madras, Chennai → Hyderabad, India 🇮🇳"
    email       = "ruthvikworking@gmail.com"

    education   = "B.S. @ IIT Madras  (Final Year)"
    roles       = [
                    "AI/ML Engineer       @ Cyepro Solutions",
                    "Founding ML Engineer @ Visionary EdTech",
                    "Tech Lead            @ Infin AI Club, IIT Madras",
                  ]

    domains     = ["Computer Vision", "NLP & Indic AI", "RAG Systems",
                   "Edge AI", "Audio ML", "ML Systems & MLOps"]

    currently_building = {
        "🌐 Visionary"    : "Multilingual RAG pipeline — 12+ Indic languages, zero LangChain",
        "🚗 Cyepro"       : "CRM-native lead scoring for automobile industry",
        "📐 Side Project" : "Multi-agent math tutoring AI for JEE-level students",
    }

    philosophy  = "Build AI that works for the next billion — not just in English."
    open_to     = ["Founding ML roles", "AI research collabs", "Competitive ML"]

`> currently_building`

🌐 Multilingual Indic RAG

@ Visionary EdTech

A production RAG pipeline serving 12+ Indic languages with zero framework overhead.

No LangChain · LlamaIndex · LangGraph
docling → chunking → hnswlib + rank_bm25
RRF fusion for hybrid retrieval
Ollama local inference on Colab T4
GridSearchCV over chunk size & embeddings

🚗 Automobile Lead Scoring

@ Cyepro Solutions

CRM-native ML pipeline to score and prioritize leads for auto dealerships.

Feature engineering from CRM event logs
XGBoost / LightGBM scoring engine
SHAP explainability for sales teams
Stage-wise funnel probability outputs
Graph DB schema (Neo4j) for lead graph

📐 Math Mentor AI

Personal Project

JEE-level multi-agent tutoring system with multimodal input and HITL feedback.

6 agents: Memory · Parser · Router · Solver · Verifier · Explainer
Claude Vision OCR for handwritten problems
Whisper ASR for voice input
FAISS RAG + human-in-the-loop triggers
Streamlit UI · automated test suite

`> experience_timeline`

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓

🚀	2025 – Present	Founding ML Engineer _{Visionary EdTech}	Building the multilingual AI core — pure Python RAG for 12+ Indic languages, embedding optimization, hybrid retrieval at scale
🏢	2024 – Present	AI/ML Engineer _{Cyepro Solutions, Hyderabad}	CRM-native lead scoring, automobile funnel analytics, codeblast analyzer with graph database design
🧑‍💻	2024 – Present	Tech Lead _{Infin AI Club, IIT Madras}	Leading AI workshops, mentoring peers on ML projects, organizing hackathons and competitive ML events
🌾	2023 – 2024	AI/ML Engineer _{Livestockify (AgriTech)}	Vision-based poultry weight estimation, Wav2Vec2 audio classification (healthy vs sick chickens), IoT sensor fusion, investor-grade PDF reporting
🎯	2023	Organizer _{Jarvis Club, IIT Madras}	HackerRank placement prep contests · 100+ participants · end-to-end contest management

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓

`> tech_stack`

Core Languages

AI / ML

Tools & Infrastructure

`> featured_projects`

🔬 Smart Poultry Farm Monitoring System

AgriTech · Computer Vision · Audio ML · IoT

End-to-end AI system for poultry health management combining vision, sound, and sensor data.

Wav2Vec2 fine-tuned for healthy vs. sick chicken audio classification
Vision pipeline for flock weight estimation (no physical contact)
IoT sensor fusion: temperature · ammonia · humidity
ReportLab investor-grade PDF health reports
Veterinary diagnostic report generation

🛰️ Satellite-Driven Aquaculture ML

Remote Sensing · Multi-task Learning

Water quality prediction for AP carp ponds using satellite imagery.

Sentinel-2 + Landsat-9 for DO, NH₃, pH, Chl-a prediction
LightGBM / XGBoost ensemble baseline
Late-fusion multi-task neural network
Backed by IIMA Ventures · IIT Kanpur TBI

📊 RAG Parameter Optimization Pipeline

Information Retrieval · NLP · Evaluation

GridSearchCV-driven systematic optimization of a RAG pipeline over CBSE Std 8 Science content.

Sweep over chunk size, overlap, embedding model, top-k
Metrics: MRR · NDCG · Hit@K
Multi-model embedding benchmarking
Reproducible Jupyter notebooks + embedding landscape guide

🚦 Smart Traffic Management — SIH

Computer Vision · Edge AI · Smart India Hackathon

AI-based adaptive traffic signal system for Indian urban intersections.

Real-time vehicle detection and density estimation
Dynamic signal timing based on live traffic load
Optimized for edge deployment constraints

`> github_stats`

`> competitive_profiles`

_{DSA · Problem Solving}

_{ML Competitions}

_{Competitive Coding}

_Algorithms

`> what_I_believe`

┌─────────────────────────────────────────────────────────────────────┐
│                                                                     │
│   "Most AI is built for Silicon Valley.                             │
│    I want to build AI for the farmer in Andhra Pradesh,             │
│    the student in rural Bihar, the shopkeeper in Tamil Nadu.        │
│                                                                     │
│    That means multilingual. That means edge-first.                  │
│    That means it actually has to work."                             │
│                                                                     │
│                                               — Sai Ruthvik         │
└─────────────────────────────────────────────────────────────────────┘

`> let's_connect`

📍  IIT Madras, Chennai / Hyderabad, India
🎓  B.S. Final Year — IIT Madras
⏰  IST (GMT+5:30) · Usually responds within 24h

_{Built from scratch · Not a template · Powered by curiosity ☕}

Provide feedback

Saved searches

Use saved searches to filter your results more quickly