👋 Hello, I'm

Nikhil Togadiya

I build autonomous multi-agent systems, constrained LLM frameworks, and production-ready AI pipelines — from research labs to real-world robotics and manufacturing floors.

View My Work About Me

3+

AI/ML Projects

2+

Industry Partners

5+

AI Technologies

Nikhil Togadiya

About Me

Turning cutting-edge AI research into real-world impact

I'm an AI/ML Engineer currently pursuing my M.Sc. in Digital Technology at Ostfalia Hochschule in Germany, specializing in Agentic AI and Generative AI systems.

My work spans building autonomous multi-agent museum guides on Temi robots with RAG-based retrieval, engineering constrained LLM frameworks for automated test generation at Volkswagen AG, and creating end-to-end AI pipelines that convert natural language into manufacturable 3D CAD models.

I thrive at the intersection of research and production — taking complex AI concepts from paper to deployment. Whether it's fine-tuning LLMs with LoRA, orchestrating multi-agent workflows, or deploying ML services in Docker containers, I love building systems that work in the real world.

🤖

Multi-Agent AI

Autonomous systems with tool-use orchestration

🧬

Generative AI

LLM fine-tuning, constrained decoding, RAG

🏭

Production ML

Docker, FastAPI, end-to-end pipelines

🤝

Industry Partners

Volkswagen, WOBCOM, Ostfalia

Professional Experience

Building AI systems for industry and research

Werkstudent (AI/Robotics)
Ostfalia Hochschule für angewandte Wissenschaften
09/2024 – Present
📍 Wolfenbüttel, Germany
  • Architected multi-agent AI museum tour guide on 2 Temi robots with RAG-based retrieval (HuggingFace embeddings + vector DB), achieving 90%+ answer coverage with sub-1s response times via confidence-gated prompting for multilingual visitor interaction.
  • Developed pytemi, a custom Python library for Temi robot control (REST API abstraction, navigation, sensors) serving as core project infrastructure.
  • Mentored team on RAG optimization and multi-robot handover protocols, coordinating integration toward production deployment.
Multi-Agent AI RAG HuggingFace LangChain Temi Robots Python
Educational Project — Constrained LLM Test Generation
Volkswagen AG
03/2025 – 07/2025
📍 Wolfsburg, Germany
  • Proposed and implemented a two-stage constrained text generation framework combining formal grammar enforcement (.lark DSL) with DFA-based token masking (Syncode) to guarantee syntactic correctness of generated system test cases.
  • Applied parameter-efficient fine-tuning (LoRA) on LLaMA 3B and 8B models to improve semantic alignment between requirement descriptions and generated test steps, analyzing performance trade-offs across model sizes.
  • Evaluated generation quality using BLEU, SacreBLEU, and ROUGE-L against reference test cases, systematically benchmarking structural validity and semantic accuracy for autonomous driving validation tasks.
LLaMA LoRA/PEFT Constrained Decoding Syncode Formal Grammars NLP
Educational Project — Energy Time Series Forecasting
WOBCOM GmbH
03/2024 – 07/2024
📍 Wolfsburg, Germany
  • Developed and optimized deep learning models (RNN, LSTM, GRU) in TensorFlow for multivariate time-series forecasting of smart meter energy data; evaluated performance using MAE and RMSE and selected the best-performing architecture.
  • Designed and implemented an end-to-end ML pipeline, including data preprocessing, feature engineering, sequence modeling, model validation, and RESTful API deployment with FastAPI for real-time inference.
  • Containerized and deployed the ML application using Docker on a local server, enabling scalable prediction services and supporting data-driven energy management decisions.
TensorFlow LSTM Time Series FastAPI Docker Python

Personal Projects

Side projects exploring the frontiers of AI

🏭

Text2CAD

Multi-Agent LLM System for 3D CAD Generation

09/2025 – 01/2026

  • Engineered an end-to-end AI pipeline converting natural language prompts into manufacturable 3D CAD models using a 4-agent architecture (requirements, planning, code generation, verification) with automated error-correction loops.
  • Implemented self-learning design memory using ChromaDB with RAG-based retrieval, enabling reuse of successful design patterns and iterative improvement.
  • Integrated sandboxed CadQuery execution, Vision-Language Model verification, and STL/STEP export via Flask web UI and CLI.
LangChain ChromaDB RAG CadQuery Flask VLM
🏒

ML-Based Ticket Sales & Revenue Forecasting

AI-Powered Attendance & Revenue Prediction for Ice Hockey

11/2025 – 11/2025

  • Engineered an end-to-end ML pipeline to forecast game-day attendance and revenue for a professional ice hockey team using 3 seasons of real sales data, achieving R² = 0.76 with ~90% prediction accuracy.
  • Designed a custom ETL pipeline and 19 domain-specific features (weather, holidays, opponent strength, team form) from multi-format sources.
  • Deployed the trained model via a FastAPI REST API with an interactive React dashboard for scenario-based revenue forecasting and data exploration.
XGBoost scikit-learn FastAPI React Feature Engineering Python

Technical Skills

Technologies I work with daily

💻 Programming Languages

Python SQL C++

🧠 ML & Deep Learning / Data Science

PyTorch TensorFlow HuggingFace Transformers scikit-learn XGBoost Transformer Architectures

🤖 Generative AI & LLMs

LLaMA OpenAI API LoRA / PEFT Prompt Engineering Constrained Decoding Syncode / DFA Token Masking Formal Grammar DSLs (.lark) Vision-Language Models

🔗 Agentic AI & RAG

LangChain LangGraph CrewAI PydanticAI OpenAI Agents SDK Tool-Use Orchestration ChromaDB FAISS Qdrant Sentence Transformers Semantic Retrieval Model Context Protocol Agent-to-Agent Protocol Confidence-Gated Prompting

📝 NLP

Text Generation Semantic Similarity Multilingual NLP Semantic Search Tokenization Evaluation (BLEU, SacreBLEU, ROUGE-L)

🚀 Deployment & Version Control Tools

Docker FastAPI Flask Streamlit Ollama REST APIs Git Linux Jupyter GitHub

📊 Data & Libraries

Pandas NumPy Matplotlib Plotly PostgreSQL Feature Engineering

👁️ Computer Vision

OpenCV YOLO Multi-View Verification

Education

My academic journey

🎓

M.Sc. Digital Technology

Ostfalia Hochschule für angewandte Wissenschaften

📍 Wolfenbüttel, Germany  |  09/2023 – Present

🎓

Bachelor in Computer Application

S.V. Patel College of Computer Science and Business Management

📍 Surat, India  |  06/2019 – 03/2022

Get In Touch

Have questions or want to collaborate? Let's connect!

Let's Talk

I'm always interested in discussing new projects, opportunities, or just chatting about AI/ML.

Email

nikhil.togadiya@outlook.com

LinkedIn

Connect with me professionally

GitHub

Check out my code

🚀 Open to Opportunities

Currently seeking full-time roles in:

  • AI/ML Engineering
  • Agentic AI & Multi-Agent Systems
  • Generative AI & LLMs
  • Research & Development
Send Message