Nilesh Raj

founder & ai engineer @ agentnine

i build and ship ai systems, from architecture to live deployment. my toolkit is python, langgraph, crewai, langchain, fastapi and docker. i treat jwt rbac, rate limiting and observability as seriously as prompt design and model accuracy.

highlights!

some things i made because i could

AgentNine
an open-source ai agent directory. every listing carries its source repo, install steps for windows, macos and linux, the access it asks for, and what's still undocumented.
live code
OmniCrew AI
autonomous multi-agent b2b lead generation. a 3-tier llm fallback router (deepseek → mistral → gemini), redis md5 caching, and ssrf-hardened scraping.
live code
LumanGuide
an agentic rag system on a langgraph state machine, with a grade-and-rewrite retrieval cycle, faiss vector search, prompt-injection guardrails and jwt rbac.
live code

on github

what i've been building in the open

my experience

agentnine
founder & ai engineer
jun 2026 - present
varanasi, india
- Founded and shipped AgentNine (agentnine.pro), an independent directory of open-source AI agents where every listing carries its source repository, install steps for Windows, macOS and Linux, the access the agent requests, and an explicit note of what is still undocumented.
- Built the product on Next.js App Router, React 19, TypeScript and Tailwind CSS with Supabase Postgres for auth and data, deployed on Vercel and covered by a Vitest suite.
- Designed the browse experience around category filters, operating-system and API-key facets, Fuse.js full-text search, and side-by-side comparison of up to three agents with no account required.
- Defined the five-step setup verification every listing passes before publication: the exact commit reviewed, the operating systems checked, the install command, a first task actually run, and the date. Grew the catalogue past 110 listings across coding, automation, research and content agents.
- Lead product direction and technical decisions, working with a small team across product, engineering and launch.
technosavvys education technology
ai/ml trainee
feb 2026 - may 2026
remote
- Engineered a BERT-based text classifier with Hugging Face to automate sentiment tagging on 500+ student course reviews per batch, refined through iterative hyperparameter tuning.
- Deployed a real-time Streamlit dashboard backed by SQLite, cutting manual review time for mentors across multiple active courses.
- Optimised NLP preprocessing pipelines, improving baseline model accuracy and accelerating weekly feedback-report generation.
- Partnered directly with course mentors to translate raw classifier output into an actionable, non-technical dashboard view.

skill stack

agentic ai & genai

LangGraph
CrewAI
LangChain
RAG Pipelines
Prompt Engineering
Vector Search
FAISS
LangSmith
DeepSeek
Mistral AI
Gemini
OpenAI API
NVIDIA NIM

ml engineering

Python
PyTorch
Scikit-learn
Hugging Face
BERT
Transformers
NLP
Fine-tuning
Pandas
NumPy

product & web

Next.js
React
TypeScript
Tailwind CSS
Supabase
Streamlit
Vitest

infra & devops

FastAPI
Docker
Redis
AWS
PostgreSQL
MongoDB Atlas
Pydantic
Git/GitHub
CI/CD
Vercel

my education

united university
bca, artificial intelligence (ibm specialization)
Prayagraj, UP
aug 2022 - may 2025

certifications

Building LLM Applications with Prompt Engineering
NVIDIA Deep Learning Institute
2025

highlights!

AgentNine

OmniCrew AI

LumanGuide

on github

my experience

agentnine

technosavvys education technology

skill stack

✦agentic ai & genai

✦ml engineering

✦product & web

✦infra & devops