Backend & AI Engineer

Ayyappa Swamy
Thati.

Architecting, deploying, and securing scalable Generative AI applications and multi-agent systems. Building the bridge between complex data and intelligent decisions.

GitHub ↗ LinkedIn ↗ Email ↗
Ayyappa Swamy Thati — ink sketch portrait

About

I specialize in the complete lifecycle of AI-powered products — from designing high-concurrency backend architectures to managing cloud infrastructure and ensuring robust network security.

Currently engineering high-performance analytics engines at Jupiter Meta, leveraging LLMs, DuckDB, and asynchronous architectures to automate complex data processing and market research at public scale.

Strong foundation in applied mathematics and data science — proficient in statistical analysis methods including multivariate regressions, Bayesian significance testing, sentiment variance models, time series analysis, and hypothesis testing. I apply these through AI-orchestrated pipelines that autonomously select the right statistical methodology for each data type.

B.Tech in Artificial Intelligence Engineering from Amrita Vishwa Vidyapeetham, Amritapuri (2021–2025).

Experience

Where I've worked.

Artificial Intelligence Engineer & System Designer Dec 2024 — Present

Jupiter Meta · Full-Time · Hyderabad

  • Independently designed and engineered the complete backend for an adaptive AI market research platform (Data-Insighter), scaling to support a mass public client base for consumer research.
  • Built a dynamic, multi-agent analytical pipeline using FastAPI, LangGraph, Google Gemini, and DuckDB for instantaneous natural language querying and automated report generation.
  • Developed an LLM-driven conversational agent that automates consumer survey generation and integrates complex data payloads, reducing research cycle time by 95%.
  • Fully manage GCP infrastructure (Cloud Run, Kubernetes Engine), database scaling, deployment lifecycles, and custom VPS network security with IP-based fraud prevention.
  • Currently building a User Persona & CRM system — survey responses flow into a FalkorDB graph database, where cohort-based filters build continuously-evolving user personas and behavioural pattern analysis from the ongoing data stream.
ML Intern Sept 2023 — Dec 2023

Amrita Wireless Network Applications · Part-Time · Amritapuri

  • Pioneered a multi-task CNN deep learning architecture that predicts object instances, human poses, instance masks and tracks people end-to-end for disaster management UAVs.

Projects

Things I've built.

Jupiter Meta · 2025

Poseidon

AI-driven analytics platform using LangGraph multi-agent workflows that autonomously classify user intent, decompose queries into analytical sub-tasks, and select optimal statistical methodologies — multivariate regressions, Bayesian tests, sentiment variance models — without rigid templates. Features a secure sandbox execution engine that crunches raw data in an isolated environment via a high-concurrency DuckDB SQL engine, achieving sub-second query responses on large datasets.

Python FastAPI LangGraph Gemini DuckDB Statistical Analysis Sandbox

Jupiter Meta · 2025

Hercules

LLM-powered conversational agent that transforms high-level research goals into structured, professional surveys. Acts as the bridge between survey creators and the SuperJ respondent app for real-world data collection at enterprise scale.

Python FastAPI Gemini API Cloud Run

Personal Research · 2025

Data Prism

A personal research AI-powered search engine and distributed document library system. Combines Perplexity-style web scraping with a private-cloud knowledge base running 100% local AI models across 3 machines. Processes PDFs, images, and audio into a searchable knowledge graph using Neo4j, Milvus vectors (HNSW), Advanced RAG, and a local Gemma LLM — with zero data leaving the network.

Python FastAPI Milvus Neo4j Gemma Crawl4ai Advanced RAG

The Liberal Labs · 2024–2025

Pocket Sage

AI-driven pipeline for clinical documentation — transcribes medical appointments via AssemblyAI, generates structured SOAP notes, and maps doctor-patient relationships in Neo4j for fraud detection and insurance analytics.

Python LangChain Neo4j FastAPI AssemblyAI

Research · 2023

Leafy

State-of-the-art plant disease diagnostic engine using custom Residual CNN with CBAM Attention Modules and Stochastic Depth. Full-stack ML lifecycle with W&B experiment tracking and progressive resizing.

Python TensorFlow PyTorch W&B

Security · 2023

Secure Remote Password

Zero-knowledge authentication protocol in Go. Server authenticates users without ever receiving passwords. SHA256 hashing with 2048-bit modular exponentiation using Sophie Germain primes.

Go SHA256 Cryptography

Computer Vision · 2024

FaceRec Attendance

Automated attendance system using face recognition — MTCNN for detection and FaceNet (InceptionResNetV1) for 512-dimensional embedding extraction. Cosine similarity matching with real-time tracking and SQLite storage.

Python Flask MTCNN FaceNet SQLite

Reinforcement Learning · 2024

TrackMania AI Driver

Autonomous driving agent for TrackMania 2020 using deep Reinforcement Learning — trained with Soft Actor-Critic (SAC) and Randomized Ensembled Double Q-Learning for real-time vision-based control.

Python SAC REDQ TMRL
🧪

Personal Development

The Liberal Labs

An open-source GitHub organization for learning through building — well-crafted, production-quality projects across AI, computer vision, distributed systems, and beyond. Every repo is a hands-on deep-dive into cutting-edge domains, built properly and shared openly.

Explore on GitHub ↗

Skills

What I work with.

Languages & Frameworks

Python Go FastAPI Flask LangChain LangGraph

Statistical Analysis & Data Science

Linear Regression Polynomial Regression Logistic Regression Bayesian Significance Testing Multivariate Regressions Sentiment Variance Models Time Series Analysis K-Means Clustering Gaussian Mixture Models DBSCAN Hypothesis Testing Pandas NumPy Matplotlib Seaborn Plotly

Machine Learning & AI

LLMs (Gemini, OpenAI, Anthropic) CNN RNN / LSTM GAN Autoencoders Transfer Learning KNN SVM Random Forest Decision Trees Naive Bayes Reinforcement Learning (SAC, DQN, SARSA) RAG Architectures Multi-Agent Systems

Cloud & DevOps

GCP Cloud Run Kubernetes Engine VM Instances Docker VPS Security Ubuntu Server Google Marketplace

Databases

MongoDB PostgreSQL MySQL DuckDB Neo4j FalkorDB Milvus ChromaDB Zillis

Web & Data Processing

FastAPI Web Scraping (aiohttp, bs4) Crawl4ai Data Visualization Knowledge Graphs User Personas

Conferences & Talks

Feb 2026

RAI Annual Event — Technical Presenter

Delivered live technical pitches of the Data-Insighter AI platform to enterprise executives at Jio Retail Center, Mumbai (Gold Sponsor).

Apr 2023

AI in Healthcare Workshop

Led hands-on learning sessions mentoring 40+ students at Amrita School of Engineering across a 3-day workshop.

Education

B.Tech in Artificial Intelligence Engineering

Amrita Vishwa Vidyapeetham, Amritapuri, Kerala

Oct 2021 — June 2025

Sri Chaitanya Educational Institution

Kukatpally, Hyderabad

June 2009 — April 2021