Hello, I'm

Praanav
Bhowmik.

|

I build production-grade Python backends, React frontends, and LLM-powered research applications. FastAPI · RAG · Agentic AI · AWS.

View My Work

praanav.py

class PraanavBhowmik:
    name  = "Praanav Bhowmik"
    role  = "Full Stack AI Engineer"
    exp   = 5+ " years"

    stack = [
        "FastAPI", "LangChain",
        "RAG",     "LangGraph",
        "PyTorch", "React",
        "AWS",     "Docker",
    ]

    def say_hello(self):
        return "Let's build something!"

01. About Me

Who I Am

Hello! I'm Praanav, a Full Stack AI Engineer with 5+ years of experience building production Python backends, React frontends, and LLM-powered research applications. I hold an M.S. in Computer & Information Science from the University at Buffalo.

My work centers on FastAPI, async APIs, RAG, agentic LLM architectures, vector search, and embedding pipelines — turning complex domain requirements into intuitive, user-centric interfaces.

I'm focused on accessible, production-grade AI: clean code, strong testing, and observable systems on AWS, GCP, and Azure.

5+ Years Exp.

20+ Projects

3 Cloud Platforms

Technologies I work with

Python
FastAPI
React
TypeScript
LangChain
LangGraph
RAG
pgvector
PostgreSQL
AWS
GCP
Docker
PyTorch
OpenAI
Claude
Gemini

02. Experience

Where I've Worked

Feb 2025 — Present Current

Research AI Engineer

@ University at Buffalo

Built production full-stack ML applications on GCP with Python (FastAPI) and async pipeline architecture; scalable REST APIs serving LLM-powered outputs, improving analytical decision-making by ~15%
Developed LLM-integrated inference pipelines with structured outputs, prompt engineering, and robust error handling using OpenAI and Hugging Face Transformers
Implemented containerised Docker AI services with CI/CD for production-reliable applications across the ML lifecycle on AWS
Led technical design reviews with cross-functional stakeholders — translating scientific requirements into pragmatic solutions

PythonFastAPIGCPAWS DockerCI/CDOpenAIHugging Face

Sept 2023 — May 2024

AI Engineer

@ Futuristic Bots

Engineered production ML inference services on AWS with containerised microservices — graph-level model evaluation, op validation, and async-optimised pipelines achieving 98% operational accuracy
Built agentic LLM workflows and event-driven data enrichment with modular, dependency-injected Python components
Owned end-to-end CI/CD: build, scan, test, Docker packaging, and deployment via GitHub Actions
Partnered with engineering and domain teams to integrate AI outputs into enterprise systems

PythonAWSDockerGitHub Actions MicroservicesAgentic LLMML Inference

Jan 2020 — Aug 2023

Junior Backend Engineer

@ Nashtech

Developed NLP and transformer-based inference pipelines with Python, PyTorch, and Hugging Face — 90%+ accuracy, ~15% fewer false positives
Built RAG systems and retrieval pipelines with vector databases, embeddings, and semantic search; shipped FastAPI services for research workflows
Designed automated ETL with Python and SQL, containerised services, and CI/CD — cutting manual processing by ~40%
Integrated OpenAI and Anthropic Claude into production with provider-agnostic routing and structured-output validation

PythonFastAPIPyTorchRAG PostgreSQLOpenAIAnthropicNLP

03. Skills

My Toolkit

Back-End & APIs

Python 3.11+FastAPIAsync I/O REST APIsPydanticSQLAlchemy PostgreSQLMongoDB

Frontend

React.jsTypeScript JavaScriptGraphQL

LLM & Agentic AI

OpenAI GPT-4Anthropic ClaudeGoogle Gemini Hugging FaceLangChainLangGraph Semantic KernelRAGPrompt Engineering

Vector & Retrieval

Azure AI SearchpgvectorChroma Semantic SearchEmbeddings

Cloud & DevOps

AWS (EC2, S3, RDS)GCPAzure DockerCI/CDGitHub Actions JenkinsOpenTelemetry

ML & Data

PyTorchscikit-learnDeep Learning NLPComputer VisionPySparkPandas

04. Projects

Things I've Built

VisualVerse

Multimodal AI platform using Python, deep learning, computer vision, NLP, and LLMs to generate contextual artwork captions — enhancing art interpretation and accessibility for ed-tech and museums.

PythonDeep LearningCVNLPLLMs

Stock Research Agentic System

Multi-agent, LLM-powered stock analysis using LangChain and RAG. Real-time pipelines integrating financial news and market data — automated investment reports improving decision-making by 20%.

PythonLangChainRAGLLMsAgentic

More on GitHub

Explore more projects on GitHub — constantly building new tools and contributing to open-source initiatives that push the boundaries of AI and software engineering.

Open SourcePythonAI

PraanavBhowmik.

|