🔗 From Next-Token Prediction to Reasoning Machines...
-
🧠 Building Finance Lab, an agentic RAG to help financial analysts avoid reading 100s of pages of SEC fillings. It uses different chunking and retriveal strategies to give you the best RAG pipeline to talk with your pdfs like 10Ks, 10Qs, merger acquisitions, etc and extract structured data like KPIs. annual revenues, etc in seconds.
-
🧠 Working on a multi agentic Resume verifier tool thats uses multiple agents working in sync to check writing quality, verify work experience, education, and other things and give a final verdict for recruiters discarding fake/fluff CVs to maker HRs life easily.
- 👨💻 I’m a Machine Learning Engineer based in London, UK.
- 🔭 I build ML & Deep Learning models from research to production.
- 🚀 Currently building Agentic RAG and compound AI systems.
- 🚀 Passionate about improving LLM reasoning, achieving SOTA results with small models and leveraging RL techniques.
- 💡 I share my learnings on tech blogs. Check out my work on Medium.
- 🤝 Passionate about contributing to Open Source.
- Masters in Applied Data Science (2023-2024)
Royal Holloway, University of London (Distinction)
- Title: Video Action-Recognition for Crowd Surveillance
- Overview: Implemented a deep learning solution combining ResNet-152, LSTM, and an Attention Mechanism to perform video action-recognition for crowd surveillance.
- Achievements: Achieved 97% accuracy, demonstrating the efficacy of integrating CNNs with temporal models for real-world surveillance challenges.
Github Repo