Looking for a job

Vaibhavi Thakur

@vaibhavithakur

I'm currently working as an SDE Intern at DataDias building backend and AI systems for voice assistants through Fine Tuning and RAG pipelines.

India

Message

What I'm looking for

I'm looking for opportunities to build reliable backend and AI systems, especially involving LLMs, RAG, machine learning, data platforms, and real-world software deployment. I'm open to remote work or relocation.

I'm building backend services for DataDias Technology Solutions' B2B voice assistant platform, integrating third-party LLMs, speech APIs, RAG pipelines, and RESTful client integrations.

At IIT Goa, I designed ensemble time-series models for stock price prediction, engineered financial data pipelines, and applied portfolio optimization to balance return and risk. At NIT Rourkela, I built a multimodal fatigue detection pipeline using speech and thermal vision that achieved 97.4% ± 3.8 accuracy under LOSO evaluation.

I've also productionized an ML-powered data quality and lineage auditor with FastAPI, PostgreSQL, Docker, React, CI-validated benchmarks, and drift-monitoring dashboards.

Experience

Work history, roles, and key accomplishments

Current

Software Engineer Intern

Current

DataDias Technology Solutions Pvt. Ltd.

Aug 2026 - Present (3 months)

Building backend services for a B2B voice assistant platform, integrating third-party LLMs and speech APIs. Implementing RAG pipelines and LLM workflows for domain-specific voice assistants across client use cases.

fastAPI Python LLM RAG REST APIs Speech to Text Text To Speech

Research Intern

Indian Institute of Technology Goa

Feb 2026 - Jun 2026 (4 months)

Designed and implemented ensemble time series models for stock price prediction using statistical and ML approaches. Evaluated model performance using quantitative metrics and conducted comparative analysis across multiple architectures.

Time Series Analysis Machine Learning Python Data Pipelines Portfolio Optimization

Summer Research Intern

National Institute of Technology Rourkela

May 2025 - Aug 2025 (3 months)

Built a reproducible multimodal fatigue detection pipeline (speech + thermal vision), achieving 97.4% ± 3.8 accuracy under Leave-One-Subject-Out (LOSO) evaluation. Developed modality-specific deep models and fusion strategies, and co-authored a research paper under active submission.

PyTorch Computer Vision MobileNetV3 Optuna Research