Skip to content
VD
VD

Berlin, Germany
Open to AI, ML, Data Science and Data Engineering roles in Germany and the EU

Vinayak Donawad

AI & Data Engineer, working across LLMs & RAG, Machine Learning, Data Analytics and Data Engineering.

I build LLM and RAG applications, machine-learning models and data pipelines. Before my M.Sc. in Data Science, I spent nearly two years as a software engineer at HCLTech shipping conversational AI to enterprise clients.

Explore my projectsRequest my resume

Ask my portfolio

BM25 retrieval, runs in your browser

Retrieving…

shipping conversational AI at HCLTech
2022–24
enterprise clients, incl. Barclays and Honda
15+
M.Sc. thesis accepted for publication
IEEE Access
projects with code on GitHub
12

About

Software engineer turned data scientist, focused on AI that holds up in production.

I started in electronics engineering, moved into software, and found the work I enjoy most where software meets data: systems that understand language, learn from data, and stay dependable once real users rely on them.

At HCLTech I built end-to-end conversational AI on the Lucy platform — conversation design, NLP intent and entity models, backend logic and production deployments — for more than 15 enterprise clients including Barclays, Honda and KONE. Integrating LLMs through prompt engineering improved response accuracy by 25%, and I connected the bots to ServiceNow and Microsoft Teams for IT service management.

My M.Sc. in Data Science in Berlin gave that experience a stronger statistical foundation. My thesis on reliability-aware retrieval-augmented generation for medical question answering has been accepted for publication in IEEE Access — and taught me to be precise about what a model can and can't claim.

I'm looking for a team where I can build AI and data products end to end: designing the pipeline, evaluating models honestly, and writing software that holds up in production.

How I work

Evaluate before you claim
I pick metrics that fit the problem — recall on imbalanced fraud data, precision for spam — and report limitations alongside results.
Ship the whole thing
Models are only useful behind a working interface. I take projects from data cleaning to an API, UI or dashboard people can use.
Keep it maintainable
Clear module boundaries, tests where they matter and READMEs that let someone else run the project in minutes.
English
C1
German
A2 — actively improving

Experience

Production software for enterprise clients, before moving into data science.

  1. Graduated Sep 2026, Berlin

    M.Sc. Data Science at University of Europe for Applied Sciences

    Completed with a thesis on reliable retrieval-augmented generation, accepted in IEEE Access. More under Research.

  2. Sep 2022 – Jun 2024, India

    Software Engineer — Conversational AI at HCLTech

    Built and deployed AI-powered chatbot systems on the Lucy platform for enterprise IT service management.

    • Designed end-to-end chatbot solutions — conversation flows, NLP intent and entity models, backend logic and production-scale deployments.
    • Improved response accuracy by 25% by integrating large language models through prompt engineering.
    • Built ServiceNow and Microsoft Teams integrations so users could raise and track IT tickets directly from the chat.
    • Worked directly with 15+ enterprise clients, including Barclays, Honda and KONE, from requirements to go-live.
    • Python
    • JavaScript
    • Node.js
    • React
    • FastAPI
    • REST APIs & JWT
    • MongoDB
    • IBM Cloudant
    • Docker
    • Kubernetes
    • AWS (S3, Lambda, CloudWatch)
    • ServiceNow
    • Microsoft Teams
    • LLMs & prompt engineering

Projects

Selected work from my GitHub — NLP and RAG systems, ML models, analytics dashboards and tested software. Each card links to the code; details pages cover methods and results.

Showing 12 of 12 projects

Featured project, AI & NLP

RAG-Scheduler

A local, privacy-first assistant that answers natural-language questions about a course timetable using retrieval-augmented generation.

  • Upload a CSV; each row is embedded once with all-MiniLM-L6-v2 and indexed in HNSWLib.
  • Queries retrieve the top-k matching rows and pass them to a local LLM via Ollama as grounded context.
  • REST API with /status, /ingest and /ask; answers return the source rows they were based on.
  • React
  • Vite
  • Node.js
  • Express
  • Transformers.js
  • HNSWLib
  • Ollama
  • Multer
Read the case study Code

Research

My M.Sc. thesis, accepted for publication in IEEE Access.

IEEE Access, accepted for publication

Reliability-Aware Retrieval-Augmented Generation for Grounded Medical Question Answering: A Reproducible and Leakage-Controlled MedQuAD Study

A fully local RAG pipeline (LangChain, FAISS, all-MiniLM-L6-v2 embeddings, Mistral 7B via Ollama) evaluated with RAGAS, plus a composite reliability score built with Platt scaling, temperature scaling and conformal prediction. The study documents its limits openly: query–document leakage, no clinical validation and manually set weights.

Grounded answering, all local, no GPUQuestionMedQuADEmbedall-MiniLM-L6-v2RetrieveFAISS, top-kGenerateMistral 7B, OllamaReliability layer (thesis contribution)RAGASevaluation metricsCalibrationPlatt, temperatureConformalpredictionScorecompositeReported with its limits: query–document leakage, no clinical validation, manually set weights.

Skills

Tools I've used in shipped work, my thesis or the projects on this page.

ProgrammingLanguages I use day to day
  • Python
  • SQL
  • JavaScript
  • Java
AI, LLMs & NLPRetrieval, generation and language models
  • LangChain
  • RAG pipelines
  • Prompt engineering
  • FAISS
  • ChromaDB
  • HNSWLib
  • Sentence Transformers
  • Ollama (Mistral, Llama 2)
  • OpenAI embeddings
  • RAGAS evaluation
  • NLTK
  • TF-IDF
Machine LearningModelling and evaluation
  • scikit-learn
  • Pandas
  • NumPy
  • TensorFlow
  • PyTorch
  • Classification (SVM, Naive Bayes, LR, RF)
  • Clustering & KNN
  • Model evaluation & calibration
  • Imbalanced data
Data Analytics & BIFrom raw data to decisions
  • Power BI
  • Streamlit
  • Matplotlib
  • Seaborn
  • Excel
  • Exploratory data analysis
  • Data cleaning
  • Time-series forecasting
Data Engineering & CloudStorage, processing and infrastructure
  • AWS (S3, Lambda, CloudWatch)
  • Apache Spark
  • Databricks
  • Snowflake
  • PostgreSQL
  • MongoDB
  • IBM Cloudant
  • Docker
  • Kubernetes
  • Splunk
Software DevelopmentAPIs, interfaces and testing
  • React
  • Node.js
  • Express
  • FastAPI
  • Spring Boot
  • REST APIs & JWT
  • pytest & unittest.mock
  • Git
  • Tailwind CSS

Education

Degrees and certifications.

  • M.Sc. Data Science

    University of Europe for Applied Sciences, Berlin, Germany — Graduated Sep 2026

    • Thesis: Reliability-Aware Retrieval-Augmented Generation for Grounded Medical Question Answering — accepted in IEEE Access.
    • Coursework projects in machine learning, decision support systems and data visualization.

  • B.E. Electronics and Communication Engineering

    KLE Technological University, Hubballi, India

    • GPA 8.25 / 10
    • Digital design, VLSI and CMOS fundamentals.

Certifications

  • AWS Certified Data EngineerAmazon Web Services
  • Google Data Analytics Professional CertificateGoogle

GitHub activity

Live from the GitHub API, refreshed each visit.

Resume

A one-page summary of my experience, education and skills.

Vinayak Shivappa Donawad — Resume

I tailor my resume to each role — ask and I'll send the latest version.

Request my resume

Contact

Open to AI, ML, Data Science and Data Engineering roles in Germany and the EU. The quickest way to reach me is email.

vinayakdon25@gmail.com LinkedIn GitHub
Open in my email app

This form opens your email app with your message filled in.