ANISH SHIRODKAR

M.S. CS · AI Engineering · Rutgers

anishshirodkar.me · v2026000

About Me

Anish

Anish Shirodkar

AI / ML Engineer · MS CS @ Rutgers

Open to Internships · Summer 2026

Building Machines
that Understand the World.

MS CS candidate at Rutgers specialising in ML, NLP, and AI engineering. I bridge research and production — shipping everything from computer vision pipelines (YOLOv8, ByteTrack) to RAG-based NLP systems (LLaMA 3.1, ChromaDB) and full-stack LLM applications. Government-registered copyright holder for an attention-LSTM weather forecasting system.

PyTorchLangChainNext.jsFastAPISupabaseDocker
Location

New Jersey, USA

ET

Lat

40.7128° N

Long

74.0060° W

Education

MSCS @ Rutgers

Major in AI · GPA 3.50

TCET · Mumbai

B.Tech IoT · CGPA 9.50 · Rank 2

Scientific Recognition

01 Copyright Registered

IP India · Cert. LD-20250175526 · 2025

0

Production AI Systems

0.00

GPA — MSCS Rutgers

0.00

CGPA — B.Tech (Rank 2)

00

Govt. Copyright · 2025

My Journey

From engineering foundations to AI research — the milestones that shaped how I build.

2026
Active

Jan 2026 – Present

Remote

Google Summer of Code

Applicant · Proposal Submitted

Submitted proposals for OLMo-7B LLM integration into DeepChem's molecular property prediction pipeline and probabilistic weather forecasting with neural-lam for MLLAM. Opened PR #4876 on DeepChem.

DeepChemOLMo-7Bneural-lamMLLAM
2025
Current

Sep 2025 – Present

New Brunswick, NJ

Rutgers University

M.S. Computer Science

Specializing in ML, NLP, and AI Engineering. Building RAG pipelines, agentic systems, and production-ready LLM applications from model to deployment.

GPA 3.50ML / NLPAgentic SystemsAI Engineering
2023
© Registered

Apr 2023 – Jun 2025

Gujarat, India

Indian Meteorological Dept.

ML Project Member · Govt. of India

Led end-to-end ML model development for a live government weather forecasting system using attention-LSTM architectures. Work registered as copyright (IP India, No. LD-20250175526, 2025).

Attention-LSTMIoT PipelineIP India 2025Production
2024
Industry

Apr – May 2024

Mumbai, India

CognoRise InfoTech

Artificial Intelligence Intern

Developed and optimized ML models for production. Conducted experiments on preprocessing strategies and feature engineering pipelines using NumPy and Pandas.

Production MLFeature EngineeringPython
2021
Foundation

Dec 2021 – May 2025

University of Mumbai

TCET Mumbai

B.Tech — Internet of Things

Built foundations in algorithms, IoT systems, and full-stack development. Achieved CGPA of 9.50/10, ranked 2nd in the batch while leading technical initiatives.

CGPA 9.50/10Rank 2IoT SystemsB.Tech
PythonPyTorchTensorFlowLangChainRAGYOLOv8Next.js 16FastAPISupabaseDockerCUDAHuggingFaceChromaDBNumPyLLaMA 3.1Gemini 2.5Attention-LSTMOpenCVPandasRedisByteTrackRL / PPOAuth0TypeScriptPythonPyTorchTensorFlowLangChainRAGYOLOv8Next.js 16FastAPISupabaseDockerCUDAHuggingFaceChromaDBNumPyLLaMA 3.1Gemini 2.5Attention-LSTMOpenCVPandasRedisByteTrackRL / PPOAuth0TypeScript
PythonPyTorchTensorFlowLangChainRAGYOLOv8Next.js 16FastAPISupabaseDockerCUDAHuggingFaceChromaDBNumPyLLaMA 3.1Gemini 2.5Attention-LSTMOpenCVPandasRedisByteTrackRL / PPOAuth0TypeScriptPythonPyTorchTensorFlowLangChainRAGYOLOv8Next.js 16FastAPISupabaseDockerCUDAHuggingFaceChromaDBNumPyLLaMA 3.1Gemini 2.5Attention-LSTMOpenCVPandasRedisByteTrackRL / PPOAuth0TypeScript

My Skills

From model training to production deployment — the full stack of technologies I use to build intelligent systems.

Hover · tap to reveal proficiency

Languages
5 skills
Python
92%
Python
JavaScript
85%
JavaScript
TypeScript
82%
TypeScript
C
72%
C
SQL
80%
SQL
AI & Machine Learning
8 skills
PyTorch
85%
PyTorch
TensorFlow
78%
TensorFlow
Scikit-Learn
85%
Scikit-Learn
Pandas
90%
Pandas
NumPy
88%
NumPy
HuggingFace
82%
HuggingFace
78%
Fine-tuning
75%
PPO / RL
Generative AI & NLP
5 skills
LLMs / RAG
88%
LLMs / RAG
LangChain
85%
LangChain
ChromaDB
80%
ChromaDB
Gemini API
82%
Gemini API
Groq / LLaMA
78%
Groq / LLaMA
Computer Vision
4 skills
YOLOv8
80%
YOLOv8
OpenCV
82%
OpenCV
75%
ByteTrack
Ultralytics
78%
Ultralytics
Web & Backend
9 skills
Next.js
85%
Next.js
React
83%
React
FastAPI
85%
FastAPI
Flask
78%
Flask
Streamlit
80%
Streamlit
Supabase
78%
Supabase
Firebase
75%
Firebase
Tailwind
88%
Tailwind
shadcn/ui
80%
shadcn/ui
MLOps & Infrastructure
8 skills
Docker
72%
Docker
AWS
70%
AWS
GCP
68%
GCP
Git
92%
Git
Vercel
85%
Vercel
W&B
75%
W&B
HF Spaces
78%
HF Spaces
Anaconda
88%
Anaconda

Featured Projects

01
COMPUTER VISION

Vtrack

Multi-class object detection pipeline (YOLOv8, ByteTrack) achieving 25-30 FPS with real-time speed estimation and 3D dashboard integration.

YOLOv8BYTETRACKFASTAPIR3FSUPABASE
02
AI ASSISTANT

SkillGap

AI career simulator using Gemini 2.5 Flash for multi-turn technical interviews and structured performance evaluation based on resume-job gaps.

NEXT.JS 15GEMINI 2.5SUPABASEPDF.JS
03
NLP / RAG

DocPilot

RAG pipeline for millisecond-latency Q&A over 500+ documentation pages, utilizing Llama 3.1 and ChromaDB for hallucination-reduced inference.

LLAMA 3.1CHROMADBHUGGINGFACESTREAMLIT
04
QUANT FINANCE

QuantVision

Backtesting pipeline that generates trading signals, evaluates cumulative returns vs buy-and-hold, and tunes parameters for strategy optimization across historical price data.

PYTHONPANDASNUMPYMATPLOTLIBJUPYTER
05
AI SECURITY

Vouch

Security control plane for AI agent authorization — policy-based access via .vouch.yml, M2M authentication through Auth0, and human-in-the-loop approval workflows for GitHub and Linear actions.

REACT 18NODE.JSAUTH0GROQ SDKEXPRESS
06
AI AGENT

VeritasAI

News aggregator AI agent that transforms noisy headline streams into an editorial front-page experience — lead stories, sentiment analysis, topic ranking, and interactive insight visualizations.

PYTHONFASTAPIPLOTLYGOOGLE NEWS RSS

Can You Beat My Agent?

Experience reinforcement learning in action. Survival is the metric - interception is the agent's only goal.

Simulation Time

0.00s

Sector Record

0.00s

Threat Level

Low

Can You Beat My Agent?

Testing predictive pursuit algorithms in a contracting spatial arena. Agent difficulty escalates exponentially after 10s.