💬Data is how we listen to reality, but data science is how we translate what we hear into something we can understand💡

About Me

• 🌱 An MBA finance and engineering graduate with a passion for addressing pressing global challenges and building innovative & sustainable AI solutions.
• 😎 Proficient in designing and deploying end-to-end ML architecture using Cloud & MLOps.

🔘 Get in touch with me 👇 🔘

📁GitHub	✉️akash.mathur2289@gmail.com
✍️Medium	🏢LinkedIn
#️⃣Kaggle	🤗Space

🛠️Generative AI Applications 👨🏽‍💻

---

Professional Experience

S&P Global – Mobility, Gurugram, India

Senior Data Scientist – Automotive Supply Chain Technology & Aftermarket (Jun 2021 – Present)

• Generated $500k+ in revenue by architecting end-to-end product on AWS that forecast Vehicles Scrappage. Setup Feature store and model monitoring with SageMaker to track and analyze metrics.
• Automated model deployment and serving with FastAPI through an efficient CI/CD integrated serverless workflow using AWS, Docker, Terraform and GitHub.
• Reduced daily news processing time by 80% by architecting end-to-end application on AWS. It recommends top news articles through similarity search and summaries generated from a fine-tuned FLAN-T5 model.
• Improved customer satisfaction and product quality by employing sentiment analysis using a fine-tuned RoBERTa model and identifying prevalent issues using topic modelling via BERTopic on AWS.
• Building a RAG system utilizing Hybrid Search and Reranking techniques using open source LLMs, while implementing fine-tuning strategies to optimize both ingestion and inferencing for enhanced performance.

Bank of America – Gurugram, India

Assistant Manager - Quantitative Modelling (Nov 2018 – Jun 2021)

• Improved bank’s risk and capital management by forecasting Equities Trading Account Profit revenue using Linear Regression as a part of regulatory compliance.
• Reduced monthly processing time by 90% by designing ETL pipelines tailored for large-scale data and leveraging ARIMA for revenue forecasting using Databricks.
• Demonstrated proficiency in Information Extraction from disparate sources and utilizing advanced analytical techniques and algorithms to make it actionable for stakeholders.

CREDIT SUISSE – Pune, India

Client Finance Analytics - Global Markets Technology (May 2015 – Nov 2018)

• Enhanced the bank’s risk management practices by designing ETL pipelines, enabling precise identification of high-risk hedge fund investors with the Random Forest classifier.
• Optimized product offerings by 40% through segmentation using Dimensionality Reduction and Clustering Techniques to identify distinct client groups, leading to deeper customer understanding.

---

Generative AI Applications

Repository

• Developed versatile applications powered by Large Language Model (LLM), including conversational AI chatbot, semantic search and text summarization, coupled with UI creation using LangChain and LlamaIndex.
• Implemented fine-tuning strategies in RAG applications, optimizing both ingestion (chunking, metadata, multi-indexing) and inferencing (query transformations, advanced retrieval strategies, re-ranking) stages for enhanced performance.
  

1. Conversational AI (Chatbot 🤖) with RAG using LangChain🦜️ & LlamaIndex🦙

• Comprehensive Guide to Customize your Llama2 ChatBot using LlamaIndex and Streamlit [Medium, GitHub]
• Mistral-7B-Instruct Based Multi-PDFs ChatBot using LangChain and Streamlit [GitHub]
• Llama2-7B Based CSV ChatBot using LangChain [GitHub]
• Simple Chatbot using AWS Bedrock [GitHub]
  

2. Advanced RAG📚📒

• Optimizing Retrieval with Additional Context & MetaData using LlamaIndex [Medium, GitHub]
• Enhancing Retrieval Efficiency through Evaluating Reranker Models using LlamaIndex [Medium, GitHub]
• Query Augmentation for Next-Level Search using LlamaIndex [Medium, GitHub]
  
• Smart Tracking and Debugging of Document Changes using LlamaIndex [Medium, GitHub]
  

3. LLM Fine Tuning 🧠 and Applications 💡

• Elevating Mistral-7B’s Performance through Finetuning using QLoRA [Medium, GitHub]
• T5 Fine Tuning & Evaluation for Text Summarization [GitHub]
• Falcon-7B Based Video 🎬 Summarization using Langchain [GitHub]
• 🎵Audio Generation🎹 using Audio Craft [GitHub]
  

4. Model Serving and Deployment ☁️

• Serverless Magic with Lambda, SageMaker DLC, and API Gateway [Medium, GitHub]
• Deploy a Serverless ML Inference using FastAPI, AWS Lambda, and API Gateway [Medium], [GitHub]
• Vector Indexing and ANN using FAISS with AWS Serverless Architecture [Medium, GitHub]

5. AWS Serverless Architectures for Big Data Workloads📊

• Optimizing Concurrent Batch Jobs Execution [Medium, GitHub]
• Effortless Scalability using Asynchronous Lambda Invocation [Medium, GitHub]
• Big Data Processing Using Distributed Maps and Step Functions [Medium, GitHub]

---

Technical Skills

🔝 Technologies

🔝 Interests

• Languages: Python, SQL, Spark • Machine Learning: Scikit-Learn, PyTorch, LangChain, LlamaIndex, Faiss, NumPy, Pandas, MLflow, TensorFlow, NLTK, spaCy, OpenCV • Cloud: AWS (SageMaker, Lambda, Batch, Step Functions, ECS, ECR, API Gateway, S3), Azure (Machine Learning, Data Factory, Databricks), • MLOps: BentoML, MLflow, DVC, Docker, Terraform, CI/CD, Git, GitHub Actions • Visualization: Tableau, Power BI • Database: MySQL, SQLite

• Generative AI • NLP • Recommendation Systems • MLOps • Stock Market Analysis, Macroeconomics • Swimming, Yoga, Meditation, Cooking

Education

• Post Graduate Diploma in Management - Financial Services

  K J Somaiya Institute of Management, Mumbai, India

  2013-2015

• Bachelor of Technology - Electronics & Communication

  Rajasthan Technical University, India

  2008-2012