t
tusharsharma173

Tushar Sharma

@tusharsharma173

Gen AI Engineer

Índia
Inglês
Algumas informações são exibidas no idioma inglês.
Sobre mim
Data Science and AI professional with 4+ years of experience in Machine Learning, Generative AI, NLP, and LLM applications. Skilled in Python, SQL, RAG, LangChain, LangGraph, Transformers, PyTorch, TensorFlow, AWS, and Databricks. Built AI solutions including Agentic RAG, conversational chatbots, PDF Q&A systems, Text-to-Speech, speech emotion detection, and retail forecasting. Improved TTS naturalness by 7%, achieved 84% accuracy in speech emotion detection, and reduced forecasting MAPE from 22% to 11%.... Saiba mais

Habilidades

t
tusharsharma173
Tushar Sharma
offline • 

Conheça meus serviços

Implementação e Implantação de IA
I will build custom ai chatbot, rag system, llm application and ai agents
A partir deUS$ 150 /projetoMais detalhes

Portfólio

Experiência profissional

Upwork

Data Scientist

Upwork • Freelance

Apr 2023 - Present • 3 yrs 6 mos

Data Scientist — Techfour Engineering Solutions Pvt. Ltd. Feb 2025 – Present Develop AI/ML and Generative AI solutions. Work on RAG, LLMs, AI agents, NLP, computer vision, and intelligent automation. Build production-oriented AI systems and document intelligence solutions. GenAI / AI Engineer — Black Basil Technologies Apr 2023 – Feb 2025 Developed Generative AI, NLP, speech AI, and conversational AI solutions. Worked with LLMs, LangChain, LangGraph, RAG, PyTorch, TensorFlow, and AWS. Key Achievements Improved Text-to-Speech speech naturalness by 7% MOS using Glow-TTS + HiFi-GAN. Built a Speech Emotion Detection model achieving 84% accuracy using Transformers. Improved retail sales forecasting MAPE from 22% to 11% using XGBoost and time-series feature engineering on 50M+ transactions. Built an intelligent PDF Q&A system processing 25K+ pages of technical and regulatory documents. Received Innovator of the Quarter award for contributions to speech projects. Major Projects Agentic RAG & Conversational AI — Built LangChain-based RAG and SQL-agent systems with multi-step reasoning, tool use, and self-reflection. Enterprise Document Intelligence / PDF Q&A — Developed OCR and non-OCR pipelines for complex documents, tables, images, and multi-column layouts with vector retrieval and LLMs. HRMS Policy & Data Agent — Created an AI system combining natural-language SQL with policy Q&A for employee, attendance, leave, and HR data. Retail Sales Forecasting — Built a Databricks/PySpark ML pipeline processing 50M+ transactions and improved MAPE from 22% to 11%. CAD Drawing Area Analysis — Developing an OCR/computer-vision solution to automatically detect rooms/zones and calculate area from architectural drawings. Text-to-Speech — Developed an end-to-end Glow-TTS + HiFi-GAN pipeline with scalable AWS training and audio-quality analysis. Speech Emotion Detection — Developed a Transformer-based speech emotion model with 84% accuracy and supporting annotation/visualization tools.