Vansh A.

Key Skills

Software

Other

Top Subject Matter

Nlp Domain Expertise

Rag Domain Expertise

LLM evaluation for QA

Top Data Types

Text

Audio

Top Task Types

Question Answering

Audio Recording

Emotion Recognition

Geocoding

Freelancer Overview

LLM vs RAG – Comparative QA System (Python, Transformers, FAISS, NLP). Core strengths include Other. Education includes Bachelor of Technology, Bennett University (2023) and Class XII, Salwan Public School (2022). AI-training focus includes data types such as Text, Audio, and Geospatial and labeling workflows including Question Answering, Audio Recording, and Emotion Recognition.

Labeling Experience

Sahaya – Geofencing-Based Tourist Safety Platform (Python, REST APIs, Machine Learning)

OtherGeocoding

Implemented geofencing-based logic to process real-time geolocation updates and trigger alerts based on whether a coordinate falls within a defined boundary. Built RESTful services to handle location updates, alert triggers, and communication between frontend and backend. Developed core backend logic in Python to evaluate point-in-region conditions for geofencing events. • Geolocation boundary definition using latitude/longitude regions • Point-in-region checks for geofencing alert triggers • Backend REST API implementation for location update handling • Alert-trigger workflow integration across system components

2025 - 2025

MarketPulse – Market Analytics Platform (Python, NLP, Machine Learning, FastAPI)

OtherTextEmotion Recognition

Built a real-time financial sentiment analytics platform that aggregates market news from multiple sources and applies NLP-based sentiment analysis for trend prediction. Implemented a FastAPI backend to deliver processed financial sentiment data efficiently in real time. Integrated transformer-based NLP models to generate Buy/Hold/Sell sentiment signals from financial news. • NLP sentiment extraction from financial news text • Transformer-based classification into market sentiment signals • Backend processing and real-time delivery via FastAPI • Actionable signal generation (Buy/Hold/Sell) from text inputs

2025 - 2025

Real time speech enhancement system (Python, PyTorch, Deep Learning, Streamlit, PyAudio)

OtherAudioAudio Recording

Developed a real-time speech enhancement system using DeepFilterNet3 and PyTorch with low-latency denoising over short audio frames. Implemented CLI, Streamlit, and live microphone inference interfaces to run streaming audio denoising with deep learning models. Continuous streaming inference reduced background disturbances in live microphone input, effectively improving audio quality in a production-like pipeline. • Real-time streaming audio inference on 20 ms frames • Speech denoising signal processing workflow • Interface implementation for live microphone processing • Performance-focused iteration to reduce audible background noise

2025 - 2025

LLM vs RAG – Comparative QA System (Python, Transformers, FAISS, NLP)

OtherTextQuestion Answering

Built and compared LLM, RAG, and hybrid QA pipelines by fine-tuning TinyLlama using LoRA, implementing FAISS-based semantic retrieval, and designing evaluation metrics to compare performance and efficiency. Encoded documents into embeddings and performed similarity-based retrieval to supply relevant context for response generation in the RAG pipeline, supporting downstream QA outputs. Evaluated approaches on accuracy, response quality, and latency to reduce hallucinations in the hybrid model. • Document ingestion and embedding generation for semantic search • Retrieval of relevant context to ground QA responses • Fine-tuning workflow setup for LLM behavior adjustment • Evaluation and rating of QA quality and latency

2025 - 2025

Education

Work History