I’m currently tinkering with agentic reasoning, retrieval systems, and context engineering. I read approx 5 research papers a week.
⚡ Fun fact: I've been to Mount Everest Base Camp Photobook here
- More details to come soon
2k25: Pixie
- Started working on this parallel to LMStudio and before Ollama UI launched
- Addressed needs for easy access to ollama models as a drop in behavioral replacement for OpenAI ChatGPT for privacy focused tasks requiring local processing
- Streamlit
2k24: Gig Match
- BM25F (weighted BM25) + Binary-Quantized HNSW-Indexed Hybrid Multi-stage Retrieval pipeline, Llama3.7b-based knowledge extraction + validation pipelines
- Hybrid Retrieval
- Weaviate
- Streamlit
2k24: Golden Retriever
- Langchain-based containerized full-stack RAG pipeline for querying personal emails
- Transformers
- Mongo
- Pinecone
- Flask
- Docker
- React
- Transformer trained from scratch in Keras for the purpose of translating natural language
- Transformers
- Keras
- Flask
- Docker
- React
2k22: Time-Series forecasting at Schneider Electric
- Temporal Fusion Transformer trained on large volumes of work-order data to predict demand
- Transformers
- RNNs
- Torch
2k22: Object Occlusion System
- YOLOv7-based detection and background imaging pipeline created with a focus on privacy
- CNNs
- Torch
- OpenCV
2k21: Training Image Segmentation models at Nimblebox
- MaskRCNN trained to segment manufactured products on a conveyor belt and chromatic consistency analysis
- CNNs
- Keras
2k21: Udacity Self-Driving Car Engineer Nanodegree
- Behavioral Cloning
- Lane Tracking
- Particle Filters
- Extended Kalman Filters
- Motion Planning
- CNNs
- Keras
- C++
- OpenCV
2k19: Presenting a poster on Adversarial Attacks in Autonomous Vehicles at Pycon India
- Survey conducted on various types of adversarial attacks applicable to autonomous vehicles
2k18: Training a Movie Recommendation Engine at Objectsol
- Java-based movie recommendation engine using Weka, trained on MovieLens1M
- Recommenders
- Java
- Weka