Shreshti Mittal — CS Engineer

⚙️

Backend Engineering

REST APIs, WebSockets, auth systems, session management, production deployments

FastAPINode.jsExpressJWTbcrypt

🤖

AI & LLM Agents

Agentic state machines, RAG pipelines, tool use, multi-turn memory orchestration

LangGraphLangChainRAGAutoGen

📊

Data Analytics

Aggregation pipelines, KPI dashboards, anomaly detection, data visualization

MongoDBPostgreSQLPower BIChart.js

👁️

Computer Vision

Real-time face mesh, head pose estimation, blink detection over WebSocket at 5fps

MediaPipeOpenCVsolvePnPEAR

Technical Skills

What I Work With

LLM orchestration, agentic workflows, RAG systems, and prompt engineering across production AI products.

LangGraph

LangChain

RAG Pipelines

AutoGen

Tool Use

CrewAI

LlamaIndex

OpenAI API

Anthropic API

Groq API

Prompt Engineering

Function Calling

Multi-turn Memory

Intent Detection

Production-grade backend systems — REST, WebSocket, auth, session management, and deployment.

FastAPI

Python

Node.js

Express.js

WebSocket

REST API Design

JWT Auth

bcrypt

Session Management

Webhook Design

Redis

Uvicorn

Database design, aggregation pipelines, analytics dashboards, and cloud data services.

MongoDB

PostgreSQL

Supabase

Mongoose

SQL

Row Level Security

Aggregation Pipelines

Power BI

Excel / KPIs

Chart.js

MongoDB Atlas

Data Modelling

Real-time computer vision pipelines with sub-100ms latency using facial landmark detection and pose estimation.

MediaPipe

OpenCV

Face Mesh (468pts)

solvePnP

Eye Aspect Ratio

Blink Detection

Head Pose Estimation

WebSocket Streaming

Rule-Based Analytics

PiP API

Full-stack delivery — from SPA frontends to cloud deployments, no framework required.

React

HTML / CSS / JS

Vercel

Netlify

Render

Web Speech API

Camera API

QR Code Generation

Git / GitHub

Glassmorphism UI

Responsive Design

Strong CS fundamentals applied in production — greedy algorithms, graph traversal, and system design.

Data Structures

Algorithms

Greedy Algorithms

Graph Traversal

LeetCode

Python

Java

Problem Solving

System Design

OOP

Azure DP-900 ✓

IBM Best Speaker ✓

Projects

Things I Shipped

AI Agent

Solo

AutoStream

// Social-to-Lead AI Agent · LangGraph + RAG + FastAPI

Conversational LangGraph agent with typed state graph, intent routing, zero-hallucination RAG, and full WhatsApp deployment via Meta Cloud API.

Directed state graph — conditional edges routing greeting / inquiry / high_intent nodes without context loss
RAG pipeline loading JSON knowledge base at startup → zero hallucination on pricing/feature queries
3-step sequential lead capture; Redis-backed session persistence across server restarts

LangGraphGPT-4o-miniRAGFastAPIRedis

GitHub ↗

CV + Backend

Team

Nexora

// Study Engagement Platform · MediaPipe + FastAPI + MongoDB

Real-time CV backend tracking focus via Face Mesh (468 landmarks) at 5fps over WebSocket — blink detection, head pose, fatigue index, and an AI coaching chatbot.

EAR (Soukupová & Čech) for blink detection; pitch/yaw/roll via OpenCV solvePnP
Weighted scoring engine — Focus Score (55% eye, 25% blink, 20% motion); hard gate at EAR < 35
MongoDB aggregation correlating focus trends with topic performance; Groq Llama 3.3 coaching

MediaPipeOpenCVFastAPIMongoDBWebSocket

GitHub ↗

HealthTech · Live

Live

SmartCheck-In

// Hospital Platform · React + Supabase + PostgreSQL

Universal healthcare platform with QR patient IDs, HIPAA-compliant Row Level Security, voice-first accessibility, and emergency registration for unconscious patients.

Dual interface — patient portal + hospital staff QR scanner with treatment notes workflow
Row Level Security in PostgreSQL via Supabase Auth; real-time sync across all hospital systems
Web Speech API voice input; emergency patient creation for time-critical scenarios

ReactSupabasePostgreSQLVercel

Live ↗GitHub ↗

FinTech · Live

LiveTeam

FinTrak

// Finance Dashboard · Node.js + MongoDB + Chart.js

Production-grade SaaS finance dashboard — no frontend framework — with JWT auth, animated analytics, group expense splitting via greedy settlement algorithm, and smart insights.

JWT + bcrypt auth, 7-day sessions; 3 Chart.js visualizations with custom gradients
Group expense module — greedy algorithm auto-calculates who owes whom across N members
Rule-based AI insights: savings rate alerts, overspend callouts, monthly summaries — zero latency

Node.jsExpressMongoDBChart.jsNetlify

Live ↗GitHub ↗