research & projects
WP1Discourse Analysis
WP2Comment Generation
Kickstart AIFood Insecurity
SideProjects & MSc
click to explore ↑
Shray Juneja
about me ↑
Experience
Education
~/shray-juneja $ whoami
Shray Juneja
APPLIED AI RESEARCHER · ENGD @ TU/E × WOONBEDRIJF
Building LLMs for real-world social impact. NLP for Dutch social housing, fraud detection & humanitarian AI.
PythonPyTorch HuggingFaceLangChain FastAPIGCPLoRA/QLoRA
click for full skills
contact
hackathons
hover objects to reveal interactions · click each to explore

Shray Juneja

APPLIED AI RESEARCHER · EINDHOVEN, NL

Applied AI Researcher specializing in LLMs, NLP, and Deep Learning, with hands-on experience across social housing, fraud detection, humanitarian AI, and tenant engagement optimization.


Proficient in Python, PyTorch, TensorFlow, and cloud-native MLOps (GCP). Skilled at bridging AI research and real-world impact through cross-functional collaboration.


Currently pursuing an Engineering Doctorate at TU/e in collaboration with Woonbedrijf — researching how LLMs can predict and influence clean energy choices in social housing.


Always happy to connect — whether it's about NLP research, social housing tech, or just a good conversation!

Experience

PROFESSIONAL HISTORY
Applied AI Researcher
TU/e × Woonbedrijf · Nov 2024 – Present · Eindhoven, NL
  • Developing LLM-driven solutions for energy transition in social housing via behavioural prediction and tenant engagement
  • Delivered a real-time Power BI dashboard for social media tenant response analysis
  • Built an LLM-powered prototype to simulate audience responses and optimise social media content pre-publication
Data Scientist Intern
HousingAnywhere · Feb 2024 – Jul 2024 · Rotterdam, NL
  • Enhanced fraud detection performance by 15% and reduced false positives via advanced ML techniques
  • Migrated legacy fraud detection codebase to GCP using MLOps, cutting deployment lifecycle from 3 to 1 day
  • Collaborated cross-functionally to align ML solutions with business and regulatory requirements
NLP Engineer Intern
Kickstart AI · Jul 2023 – Aug 2023 · Delft, NL
  • Improved food insecurity predictions for the Horn of Africa using topic modelling, clustering, and geocoding
  • Integrated IPC dataset with news article insights to train a robust ML classifier
  • Achieved 30% increase in prediction performance over human baselines

Education

ACADEMIC BACKGROUND
Engineering Doctorate (EngD)
Technical University of Eindhoven · 2024 – Present
Project: Large Language Model to predict and influence clean energy choices in social housing
MSc Artificial Intelligence
Rijksuniversiteit Groningen · 2021 – 2023
Thesis: Estimation of Uncertainty for Audio Classifiers
BTech Computer Science & Engineering
SRM University · 2016 – 2020 · Chennai, India

Skills

TECHNICAL STACK

AI & Machine Learning

PythonPyTorchTensorFlowHuggingFaceLLM Fine-tuningLoRA / QLoRARAGLangChainDutch NLPScikit-learn

MLOps & Cloud

Google Cloud PlatformVertex AIDockerKubernetesKubeflowJenkinsGitFastAPI

Data & Analytics

Power BILooker StudioSnowflakeMicrosoft FabricAzure

Research & Projects

ENGD · INTERNSHIP · MSC · SIDE PROJECTS

EngD Research

WP1 — Tenant Discourse Analysis on Social Media
Analysis of 3,197 Dutch housing association Facebook comments (92 associations, 2018–2023). Sentiment, intent, and post-relatedness classification via BERT ensemble (GroNLP, DeBERTa v3, RobBERT 2023). K-Means clustering to uncover latent discourse types.
→ github
WP2 — Synthetic Tenant Comment Generation
Fine-tuning Llama 3.2 3B, Gemma 4B, and Fietje 2.7 for synthetic Dutch Facebook comment generation. Comparing few-shot, objective, and multi-task approaches.
→ github

Internship

Food Insecurity Prediction — Kickstart AI
NLP pipeline for food insecurity prediction in the Horn of Africa. Topic modelling, clustering, geocoding. 30% improvement over human baselines.
→ github

MSc Thesis

Uncertainty Estimation for Audio Classifiers
Multi-label audio classification with ResNet50 and Monte Carlo dropout for uncertainty estimation and calibration analysis.
→ github

Side Projects

Airfoil Design Surrogate Modelling
NACA 2412 design space exploration. 9 regression models benchmarked with Ordinary Kriging.
→ github
Twitter Sentiment Classification (BiLSTM)
3-class tweet sentiment using BiLSTM with spaCy tokenisation.
→ github

Hackathons

BUILDING FAST · SHIPPING BOLD
Bunq Hackathon — Agents 6.0
AI agents-powered dashboard analysing spending habits, estimating carbon footprint, and encouraging climate-friendly financial decisions.
→ github
Tech Europe Hackathon 2026 — Clash Royale Analyst
ESPN-style post-match Clash Royale analysis with AI coaching. Upload match recording → broadcast-quality breakdown: deck stats, AI commentary, highlight video.
→ github

Let's connect

ALWAYS HAPPY TO CHAT

Whether it's about NLP research, social housing tech, a collaboration, or just a good conversation — feel free to reach out!

✉ shrayjuneja15@gmail.com ✆ +31-684-792-622 in LinkedIn → ⌥ GitHub →