|
●
●
●
zsh — sujith@developer-core: ~/workspace (active)
🔍 View Plain Text Terminal Output┌──(sujith㉿developer-core)-[~]
└─$ whoami
Sri Sai Sujith Yalahmanchi
┌──(sujith㉿developer-core)-[~]
└─$ role
Full-Stack Developer • AI Engineer • Agent Builder
┌──(sujith㉿developer-core)-[~]
└─$ education
B.Tech — Computer Science & Engineering @ Amrita Vishwa Vidyapeetham
┌──(sujith㉿developer-core)-[~]
└─$ focus
AI Agents • Agentic Workflows • Production RAG • Java Backend • Distributed Systems
┌──(sujith㉿developer-core)-[~]
└─$ building
Autonomous multi-tool agents, high-throughput APIs, and interactive full-stack platforms
┌──(sujith㉿developer-core)-[~]
└─$ philosophy
Build real systems → Master core mechanics → Optimize bottlenecks → Ship value |
I am a Computer Science & Engineering student at Amrita Vishwa Vidyapeetham focused on the intersection of Full-Stack Software Engineering, Java Backend Architecture, and Autonomous AI Systems.
- 🤖 Agentic Systems & RAG: Designing stateful multi-tool agents with LangGraph, cyclic self-correction loops, and dense FAISS vector retrieval systems grounded in authoritative datasets.
- ⚙️ Robust Backend Architectures: Building scalable server-side systems with Java, Spring Boot, REST APIs, and high-performance Python FastAPI services.
- 🌐 Modern Full-Stack Applications: Developing reactive, data-intensive web experiences with React, TypeScript, Tailwind CSS, and 3D WebGL visualizations with Three.js.
- 🔬 Systems & Problem Solving: Deep algorithmic background spanning graph theory, memory-conscious data structures, OS concurrency (POSIX threads/IPC), and hardware integrations.
I don't just consume abstractions — I build end-to-end architectures and ship working software.
|
Cyclic state machines with LangGraph, ReAct multi-tool reasoning, SQL generation over Parquet, and conformal abstention gates for safety-critical execution. |
End-to-end document pipelines: recursive chunking, dense vector indexing via FAISS, HuggingFace embeddings, and low-latency Groq LPU inference. |
Enterprise-ready Java & Spring Boot services, async FastAPI servers, Prisma ORMs, and secure authentication models (TOTP 2FA, JWT). |
|
Responsive single-page web applications with React, TypeScript, Vite, Tailwind CSS, Framer Motion, and 3D globe visualization in Three.js. |
In-memory analytical querying with DuckDB, tabular data wrangling with Pandas/NumPy, and comparative machine learning pipelines. |
Robotic scene-text SLAM pipelines combining EasyOCR + semantic mapping, and algorithmic drone dispatch simulations in Java. |
Agentic AI is more than prompting — it is about state orchestration, reliable tool calling, memory management, and deterministic verification.
┌─────────────────────────────────────────────────────────────────────────────┐
│ AGENTIC REASONING & TOOL ENGINE │
└─────────────────────────────────────────────────────────────────────────────┘
│
[ User Query ]
▼
┌───────────────────────────────┐
│ FASTAPI ASYNC GATEWAY │
└───────────────┬───────────────┘
│
▼
┌───────────────────────────────┐
│ LANGGRAPH REASONING AGENT │
│ • State Graph Execution │
│ • Plan & ReAct Loop │
│ • Conversational Memory │
└───────┬───────────────┬───────┘
│ │
┌───────────────┘ └──────────────┐
▼ ▼
┌─────────────────────────┐ ┌─────────────────────────┐
│ ANALYTICS & SQL │ │ SEMANTIC RETRIEVAL │
│ • DuckDB In-Memory │ │ • FAISS Vector Store │
│ • Parquet Data Lake │ │ • Sentence Embeddings │
│ • Dynamic Query Exec │ │ • Domain Document RAG │
└────────────┬────────────┘ └────────────┬────────────┘
│ │
└───────────────────────┬──────────────────────┘
│
▼
┌───────────────────────────────┐
│ SAFETY & VERIFICATION │
│ • Factual Claim Checking │
│ • Cyclic Self-Correction │
│ • Conformal Abstention Gate │
└───────────────┬───────────────┘
│
▼
┌───────────────────────────────┐
│ FINAL SYNTHESIZED OUTPUT │
└───────────────────────────────┘
| Core Frameworks | Tool Integration | Vector & Storage | Verification & Safety |
|---|---|---|---|
LangGraph • LangChain |
DuckDB SQL • FastAPI Tools |
FAISS • HuggingFace Embeddings |
Cyclic Claim Correction |
Groq LPU (Llama 3) • Python |
Zone Lookup • External APIs |
Parquet Data Stores |
Split Conformal Abstention |
Java •
Python •
C / C++ •
TypeScript •
JavaScript •
SQL •
Haskell •
MATLAB •
Bash •
HTML5 / CSS3
React 18/19 •
Next.js •
TypeScript •
JavaScript •
Vite •
Tailwind CSS •
Three.js (3D WebGL) •
Redux •
Framer Motion •
shadcn/ui
Spring Boot •
FastAPI •
Node.js •
Express.js •
Prisma ORM •
Hibernate / JPA •
GraphQL •
Flask •
RESTful APIs •
Uvicorn
PostgreSQL •
MySQL •
MongoDB •
SQLite •
Supabase •
Redis •
DynamoDB •
DuckDB (In-Memory OLAP) •
FAISS (Vector Index)
LangGraph •
LangChain •
PyTorch •
Scikit-Learn •
TensorFlow •
OpenCV •
Hugging Face •
Groq LPU API •
Sentence Transformers •
Pandas •
NumPy •
Streamlit
Agentic RAG •
Cyclic Claim Correction •
Split Conformal Prediction •
QLoRA Fine-Tuning •
Dense Semantic Search •
EasyOCR •
Computer Vision
Git •
GitHub •
GitHub Actions •
Docker •
AWS •
Vercel •
Render •
Linux •
Ubuntu •
Windows •
Wireshark •
TCP/IP Sockets
Arduino •
Figma •
Postman •
VS Code •
IntelliJ IDEA •
Eclipse •
Cursor •
Google Antigravity •
Model Context Protocol (MCP)
🤖 LP-SLAM-TECC
Enhances robotic spatial SLAM by integrating scene-text comprehension into semantic mapping. Detects environmental text with EasyOCR, eliminates OCR noise via Text Error Correction & Classification (TECC), enriches understanding with FAISS vector RAG, and builds a queryable semantic map.
|
🧠 NYC Taxi Insight Agent
Agentic reasoning system built with LangGraph to orchestrate complex analytical tasks over massive open datasets. Dispatches multi-tool ReAct loops over DuckDB for high-speed SQL analytics across millions of taxi trips, performs FAISS vector search on metadata PDFs, and logs executions.
|
🛡️ SafeRoute-CDS
Safety-critical Clinical Decision Support (CDS) architecture guarding against LLM hallucinations. Employs a compiled cyclic LangGraph state machine that extracts factual claims, validates against medical texts, self-corrects invalid assertions, and applies a Split Conformal Prediction abstention gate.
|
🌍 GlobeRadio
A full-stack platform enabling users to navigate an interactive 3D WebGL globe, click any territory, and instantly stream live radio broadcasts. Engineered with TOTP-based two-factor authentication, personal favorite feeds, and listening history tracking.
|
⚡ Groq Chatbot RAG
Intelligent document question-answering system capable of ingesting 300+ page enterprise PDFs. Deconstructs text into semantically cohesive chunks, generates dense vector representations with Sentence Transformers, indexes via FAISS, and streams sub-second answers using Groq LPU models.
|
|
🔬 SciVerify
Natural Language Processing web system verifying scientific assertions against biomedical evidence. Evaluates claim veracity by retrieving relevant research and classifying statements as Supported, Contradicted, or Not Enough Info inside a modern reactive UI.
|
🛸 DSA Drone Delivery System
Enterprise logistics simulation modeling autonomous drone fleet scheduling. Leverages foundational data structures and algorithms — graph routing, priority dispatch queues, linked structures, and load-balancing heuristics to optimize package turnaround times.
|
|
🤖 Agentic AI & Workflows LangGraph state machines, cyclic claim self-correction, ReAct loops, multi-tool calling |
🔍 RAG & Vector Engines Dense FAISS indexing, semantic search, document chunking, sub-second Groq inference |
⚙️ Java Backend & APIs Spring Boot, REST architecture, Hibernate/JPA, async FastAPI, Prisma ORM |
|
🌐 Full-Stack Web Development React, TypeScript, Vite, Tailwind CSS, 3D WebGL (Three.js), modern responsive UI |
📊 Data Infrastructure & OLAP DuckDB over Parquet, PostgreSQL, MySQL, MongoDB, SQLite, Pandas data wrangling |
🔌 Systems, OS & Networks C concurrency (pthreads, mutexes, IPC), Arduino robotics, QoS network analysis |
|
|
|
|
|
|
|
Bachelor of Technology (B.Tech) — Computer Science & Engineering
|
┌──(sujith㉿engineering-core)-[~/core]
└─$ cat philosophy.txt
[01] BUILD REAL THINGS
Software is validated by execution, not by speculation. Write code that runs.
[02] UNDERSTAND THE MECHANICS
Never be satisfied with high-level abstractions alone. Understand how the memory,
the database engine, the network sockets, and the LLM inference loops function underneath.
[03] ARCHITECT FOR RESILIENCE
Systems fail at the seams. Validate inputs, calibrate confidence gates, handle
concurrency cleanly, and design modular state machines.
[04] CONTINUOUS ITERATION
Build → Measure → Learn → Refactor → Ship. |
Looking to collaborate on agentic AI systems, scalable backend architectures, or open-source software? Let's connect.
⚡ Engineered with curiosity, precision, and clean code • Sri Sai Sujith • Always building. Always learning. ⚡