Ashish Ranjan
Development
Bengaluru, India
Skills
Machine Learning (ML)
About
Ashish Ranjan's skills align with IT R&D Professionals (Information and Communication Technology). Ashish also has skills associated with Programmers (Information and Communication Technology). Ashish Ranjan appears to be an entry-level candidate, with 4 months of experience.
View more
Work Experience
Data Scientist
Evidence Extraction
March 2024 - Present
- Freelance - Developed a Document Question Answering platform using Gen-AI to extract key evidence from Audit and compliance reports, significantly reducing manual effort. - Conducted data extraction from complex PDF documents using pdfplumber and Azure Document Intelligence - Leveraged Azure OpenAI text-embedding-ada-002 for precise vectorization and developed a Question- Answering pipeline incorporating semantic similarity and GPT-4 - Seamlessly wrapped up the code base as a Rest API endpoint using Django, reducing human intervention by 60% through automation while enhancing persistence and accuracy of generated evidence.
Data Scientist
OpenAI
November 2023 - November 2023
- Customer Experience Intelligence Kanini - Developed scalable code to produce precise transcripts from medical audio data, achieving an accuracy exceeding 98% through OpenAI Whisper. - Seamlessly aligned transcriptions with audio files utilizing a large-scale neural network, yielding aligned transcription files suitable for speaker diarization, sentiment analysis, and language detection. - Implemented Speaker Diarization using NVIDIA NeMo, accurately mapping speakers to corresponding sentences. Additionally, ensured proper punctuation in the transcriptions for clarity and readability. Data Scientist: RT Noise Cancellation Kanini - Conducted comprehensive training on diverse datasets to ensure the robustness and efficacy of the model. - Engineered Django backend infrastructure to streamline audio input processing and machine learning-based noise suppression, achieving a notable 60% enhancement in audio quality. Additionally, wrapped up the codebase as a Rest API endpoint for seamless integration with ML processes. Data Scientist: Deloitte Neuroscience Kanini - Developed scalable code for analyzing animations, font, images, word count, bullet points and content after extracting data and metadata from PowerPoint files to offer tailored improvement suggestions in collaboration with Neuroscience Research team. - Utilized BERT for Extractive summarization and OpenCV with OCR for extracting text from slide images. Employed Spacy and NLTK for Named Entity Recognition (NER) to word identification requirements. - Improved presentation analysis efficiency by threefold, garnering positive user and stakeholder feedback for increased productivity.
Data Scientist
Deloitte HOAIML
August 2022 - August 2022
- Kanini - Orchestrated automation of file uploading processes, facilitating seamless conversion to different formats and subsequent restoration post-OCR processing. - Elevated table detection accuracy from 60% to over 95% through innovative algorithms, while also revamping documentation for improved clarity and accessibility, and utilizing PowerBI and Apache Superset for data analysis and dashboard development..
SDE Intern
February 2022 - February 2022
- Conversational Assistant Kanini - Spearheaded the development of the backend infrastructure, ensuring scalability and efficiency from inception. - Trained and fine-tuned models to interpret user intents, refining responses for enhanced user interaction. - Implemented REST API deployment, rigorously tested with Postman, and seamlessly integrated with ReactJS for optimal user experience.
Student Developer
Peeple
January 2020 - January 2020
- Winter of Code 2019 - Developed a signup feature enabling image uploads and face comparison with database entries. - Established database architecture to store user information and images securely. - Seamlessly integrated Flask backend with an existing Convolutional Neural Network (CNN).
Education
