AVAILABLE FOR OPPORTUNITIES
New Delhi, India
2nd Yr · B.Tech CS (AI)

Suhaani

Agarwal

_

2nd-year CS (AI) student at IGDTUW Delhi — but honestly, the degree is the least interesting thing happening right now.

GSoC contributor turned mentor in under 12 months. 7× hackathon winner, including Smart India Hackathon 2025 — India's largest, 8 lakh+ students. Incoming summer intern at Motorq, Bangalore, after clearing one of the most competitive interview processes at my college.

I build agentic AI systems, voice-first interfaces, and production-ready LLM pipelines — not as side projects, but as real working products solving real problems. I'm always chasing what just dropped, what just became possible, and what nobody has built with it yet.

Still early. Always building.

🏆 SIH '25 Winner🔬 GSoC '25 Top 1.2%🎓 GSoC '26 Mentor · R Project⚡ 7× Hackathon Winner🏢 Motorq Intern '26🤖 Agentic AI
Resume
Python ·TypeScript ·Next.js ·LangChain ·FastAPI ·PyTorch ·React ·GSoC '25 Contributor ·GSoC '26 Mentor ·SIH '25 Winner ·Motorq Intern '26 ·7× Hackathon ·Agentic AI ·RAG ·MCP ·Docker ·PostgreSQL ·Open Source ·Python ·TypeScript ·Next.js ·LangChain ·FastAPI ·PyTorch ·React ·GSoC '25 Contributor ·GSoC '26 Mentor ·SIH '25 Winner ·Motorq Intern '26 ·7× Hackathon ·Agentic AI ·RAG ·MCP ·Docker ·PostgreSQL ·Open Source ·
02

Achievements & Experiences

10 total
MOTORQ Summer Intern 2026
SELECTED
Bangalore · Summer 2026

MOTORQ — 1 of 1 from campus

MOTORQ Summer Intern 2026

The only student selected from my entire college for Motorq's competitive internship programme — heading to Bangalore in Summer 2026. The selection process was one of the most gruelling on campus, spanning multiple elimination rounds before a final 1:1 with the CTO. Out of every student who competed, I was the sole offer.

SELECTION PIPELINE

1Resume Shortlisting
2Hard DSA Online Assessment
33-4 Hard DSA Interview Rounds
4System Design Round
5Database + OOP + CS Fundamentals
6CTO Interview
03

Selected Work

PROJECT 01
CHROME EXTENSION

ctrl

The AI That Uses AI — Voice‑First Browser Agent

Most people treat AI as a single tool. CTRL changes that. It's a Chrome extension with a voice-enabled avatar that understands what you're trying to do — then routes to the fastest possible AI tool, writes an optimised prompt, and executes the task.

When you say "make a presentation," CTRL doesn't build slides manually — it opens Gamma, generates a complete deck in one step, and returns the result. Tasks that take minutes of clicking are reduced to seconds. The avatar narrates every action in real time, making it both an execution layer and a guide to the AI ecosystem.

Voice Avatar (Rive + TTS)

A lip-synced 2D Rive avatar listens, narrates, and explains every action in real time. Feels like a conversational partner, not automation.

AI Tool Router

Picks the shortest execution path — opens Gamma for decks, Framer for design, Perplexity for research. 20+ web AI tools supported.

Chrome DevTools Protocol

Full CDP access via accessibility tree — clicks, scrolls, keypresses, navigation, form fills. More reliable than DOM-based approaches.

Gemini Live API

Low-latency streaming dialogue with interruption handling. Maintains context across tasks and asks clarifying questions when needed.

OSS LLMsGemini Live APIChromaDBChrome DevTools APIRiveMediaRecorder APINext.js
gamma.ai — New Deck
Perplexity
+3
🔒 gamma.ai/new
γ
gamma.ai— Create with AI
Generating

↓ PROMPT — AUTO-TYPED BY ctrl

Create a professional 12-slide presentation on renewable energy — solar, wind & hydro — with real data visualisations, executive summary, and modern gradient theme|

Slide 1 — Cover

Slide 2 — Overview

Slide 3–12…

Scraping energy data sources
Selecting chart types via LLM

ctrl

Voice AI Agent

LIVE

YOU SAID:

“make a presentation on renewable energy”

Intent: slide deck on renewables
Routing → gamma.ai (fastest path)
Injecting optimised prompt…
Generating 12 slides…
PROJECT 02
SIH 2025 WINNER

Metro
Mind

Kochi Metro Rail AI Platform · SIH Winner '25

Every night, KMRL's operations team spent nearly two hours manually reviewing scattered spreadsheets — train induction, standby rakes, withdrawals, manpower — all under complex constraints.

MetroMind eliminated that. A CP-SAT multi-objective optimizer explores 47,000+ scheduling possibilities — balancing maintenance, fitness certificates, energy consumption, SLA penalties, branding exposure, and costs. Tasks that took 2 hours now finish in under 5 minutes.

SCHEDULING TIME

2 hours

< 5 min

OPERATIONS METHOD

Spreadsheets

AI Engine

ROSTER GENERATION

Manual

Automated

CP-SAT Optimizer

Solves 47,000+ scheduling possibilities — maintenance, fitness certificates, energy, SLA penalties, branding hours, costs — in seconds.

WhatsApp AI Agent

Real-time team coordination via WhatsApp. Automated alerts for delays, status updates, and operator notifications.

Voice AI (ElevenLabs)

Natural-language voice calls for train operators. Ask scheduling questions conversationally — get instant answers.

Depot Parking Plan

Optimal overnight parking considering maintenance bays, washing schedules, and next-day induction order. Auto-generated every night.

CP-SATOSS LLMsElevenLabsFastAPIFirebaseNext.jsWhatsApp APIML Models
CP-SAT Optimizer — Running
✓ Solved in 3.2s

CONSTRAINTS BEING EVALUATED

Maintenance windows4 rakes due service
Fitness certificate status2 expired — skip
Energy minimisation (shunting)Paths optimised
Branding exposure hours≥ 6h/day required
SLA penalty reduction0 violations target
Cost optimisationMinimising depot idle
SCHEDULING POSSIBILITIES EXPLORED47,382

Best schedule selected — next day plan + depot parking ready

NEXT DAY — SCHEDULE + DEPOT PARKING PLAN

KL-04Aluva → ThripunithuraINDUCTEDBay 2
KL-07MG Road → AluvaSTANDBYBay 5
KL-02Edappally → Maharaja'sINDUCTEDBay 1
KL-11Thripunithura → AluvaSCHEDULEDBay 3
WhatsApp AI Agent
KL-04 delay 5 min at Edappally
Notifying 6 operators...
Voice AI (ElevenLabs)

Natural language voice call active

PROJECT SCREENSHOTS

1 / 1
Aluva
Edappally
MG Road
Maharaja's
Thripunithura
01KOCHI METROMetroMind AI
01KOCHI METROMetroMind AI
PROJECT 03
AI ANALYTICS ENGINE

viz.ai

Autonomous Interactive Dashboard Generator

Upload any raw dataset — viz.ai handles everything else. A multi-orchestration LangGraph pipeline coordinates open-source LLMs to automatically select visualization types, detect patterns and anomalies, and generate a complete multi-plot analytics dashboard.

Every chart is editable in natural language: "make the bars blue," "add a trend line," "switch to a scatter plot." Real-time rendering under 8 seconds per request via FastAPI + Next.js + Docker + Nginx.

5+

Connected visualisations per dashboard

~70%

Reduction in manual editing time

< 8s

End-to-end render time

LangGraph Orchestration

Coordinates specialized LLMs for chart selection, insight generation, and styling — all in one pipeline.

Natural Language Edits

Describe any change in plain English. The dashboard updates instantly — ~70% less manual editing time.

ML Insight Generation

Auto-detects patterns, outliers, and trends in your data — surfaces insights you might miss at a glance.

Production Deployment

FastAPI + Next.js, containerised with Docker, served via Nginx. Under 8 seconds per full dashboard render.

LangGraphOSS LLMsRAGChromaDBFastAPINext.jsDockerNginxAWS EC2
viz.ai — Dashboard Generated
✦ NL Edit:“make the revenue chart a line graph with gradient fill”↵ Apply

MONTHLY REVENUE

SEGMENT MIX

A — 59%
B — 24%
C — 17%

TREND — AUTO-DETECTED BY ML

ANOMALY SCATTER

Auto-insight: 23% revenue spike in Q3. 1 anomaly flagged in scatter.

PROJECT SCREENSHOTS

1 / 11
PROJECT 04
AI COMPLIANCE PLATFORM

vigilo

AI-driven Compliance Monitoring Platform

Regulatory compliance is a full-time job — Vigilo automates it. It continuously scrapes 5+ government portals (FSSAI, RBI, SEBI, Gazette, MCA) to detect new amendments the moment they're published, storing them as vector embeddings in ChromaDB for RAG-based retrieval.

Groq-hosted LLMs (GPT-OSS 120B, LLaMA-3.3-70B) summarise dense regulatory text into actionable insights — reducing manual review time by ~80%. Teams get automated alerts via SendGrid before deadlines hit.

5+

Government portals monitored

~80%

Reduction in manual review time

Real-time

Amendment detection & alerts

Multi-Portal Scraping

Automated scrapers for FSSAI, RBI, SEBI, Gazette, MCA — detecting new amendments instantly on publication.

RAG-Based Retrieval

ChromaDB vector store enables precise, contextual retrieval of regulations relevant to each user's compliance profile.

LLM Summarization

Groq-hosted LLMs (GPT-OSS 120B & 20B, LLaMA-3.3-70B) summarise dense regulatory text into clear, actionable reports.

Automated Alerts

SendGrid-powered email alerts with JWT-secured Next.js dashboard tracking amendment status and compliance scores.

LangChainChromaDBRAGPrompt ChainingGroq LLMsFastAPIPostgreSQLNext.jsJWTSendGrid
vigilo — Compliance Monitor
Live Scanning

PORTALS MONITORED (5+)

FSSAI2 new
RBI1 new
SEBI✓ ok
Gazette3 new
MCA1 new

LATEST AMENDMENTS DETECTED

FSSAIHIGH

Food labelling amendment 2025

Today
RBIMEDIUM

KYC norms — digital onboarding update

Yesterday
GazetteLOW

GST reconciliation deadline extended

2d ago

RAG QUERY — Natural Language

“What FSSAI changes affect my food import business this quarter?”

↳ 3 relevant amendments retrieved via ChromaDB. Key impact: new labelling requirements for imported goods effective Q2 2025. Alert sent via SendGrid.

PROJECT SCREENSHOTS

1 / 8
05+

More Projects

05
Accessibility Dashboard

Sweekaar

Unified accessibility dashboard for neurodivergent children using computer vision, object detection, and adaptive learning experiences.

Next.jsNode.jsFirebaseMediaPipe+1
06
Memory Care Platform

MemoryHaven

Platform for individuals with memory impairments — location tracking, AI-powered reminders, and personalised cognitive assistance.

Next.jsNode.jsOpenAI APIPostgreSQL+1
07
Student Wellness

AASRA

AI-powered platform for PG students addressing healthy eating and emotional well-being through personalised meal planning.

Next.jsNode.jsOpenAI APIMongoDB+1
04

Tech Stack

A snapshot of the technologies I build with — from AI pipelines and LLM systems to full-stack products shipped to production.

Languages
JavaScript
Python
TypeScript
C++
C++
R
R
AI & Machine Learning
TensorFlow
PyTorch
JUP
Jupyter
OpenAI
OSS
OSS LLMs
Agentic AI & LLMs
LAN
LangChain
LAN
LangGraph
RAG
RAG
MCP
MCP
Backend
Node.js
FastAPI
EXP
Express.js
MongoDB
PostgreSQL
Frontend
React.js
Next.js
Tailwind CSS
HTML5/CSS3
Tools & Cloud
Git
GitHub
Docker
AWS
AWS
Firebase
VS Code
Postman
05

Get in Touch

Let's build something
extraordinary!

Open to interesting projects, AI collaborations, research, and conversations. Whether you have a wild idea or a hard problem — let's talk.