Highlights
A collection of engineering challenges I have tackled recently, spanning from distributed container orchestration to high-concurrency booking engines and enterprise RAG systems.
HemGPT: Enterprise RAG System
A production-grade RAG platform enabling secure multi-document reasoning over 10,000+ internal documents via hybrid vector search, Reciprocal Rank Fusion, and local Ollama inference.
What it does
- •Engineered a hybrid retrieval engine pairing ChromaDB vector embeddings with BM25 keyword search, fused via Reciprocal Rank Fusion.
- •Implemented cross-encoder reranking and context compression to eliminate low-relevance passages before LLM context injection.
- •Integrated local Ollama inference for query expansion and token streaming with zero external API data leakage.
AuraDeploy: Distributed Container Orchestration Engine
A high-availability container orchestration engine written natively in Go, leveraging embedded Raft consensus, custom CNI networking, and CRI-O/containerd OCI runtimes.
What it does
- •Architected an Active-Passive high-availability control plane in Go using embedded HashiCorp Raft for distributed consensus, log replication, and FSM snapshots with zero external DB dependency.
- •Integrated native CRI-O / containerd runtime interfaces (`containerd/oci`) for staging OCI images, configuring cgroup limits, and managing network namespaces via `netlink`.
- •Designed a custom scheduling loop with predicate filtering (`HasSufficientResources`, `VolumeNodeAffinity`) and `LeastAllocated` priority scoring for optimal cluster workload placement.
SunX: AI Radiography & Clinical Decision Support
An AI radiography platform integrating a DenseNet-121 vision model for chest X-ray pathology detection with a pgvector RAG engine to surface evidence-based clinical treatment guidelines.
What it does
- •Built a DICOM image processing and HIPAA-aware PHI anonymization pipeline using `pydicom` and `MONAI` for automated medical image ETL.
- •Deployed a fine-tuned DenseNet-121 PyTorch vision model on NVIDIA Triton Inference Server to classify 14 pathology classes (Pneumonia, Cardiomegaly, Effusion, etc.).
- •Engineered a clinical RAG pipeline using LangChain and `pgvector` over 1536-dimensional embeddings, retrieving treatment recommendations grounded in AHA/ACC and IDSA medical guidelines.
ShuttleNow: Real-Time Shuttle Booking Platform
A real-time transit reservation engine featuring Socket.IO seat soft-locking, live Google Maps route tracking, Stripe payments, and digital QR tickets.
What it does
- •Eliminated double-booking race conditions by 95% using Socket.IO in-memory soft-locks that instantly reserve selected seats across all connected clients before database writes occur.
- •Integrated Google Maps Directions API for live route drawing and Google Places Autocomplete for dynamic admin location entry.
- •Architected secure payment verification using Stripe API webhooks and serverless functions for PCI-compliant transaction processing.
E-Commerce GraphQL Architecture
Led redesign of backend GraphQL services for a high-traffic e-commerce platform serving 50k+ daily requests at Speeler Technologies.
What it does
- •Reduced P95 latency by 30% by implementing resolver batching and AppSync caching.
- •Designed a multi-tenant DynamoDB Single-Table Architecture, utilizing sparse indexes and GSI sharding to eliminate hot partitions and support high-cardinality access patterns.
- •Engineered a fault-tolerant, event-driven pipeline using Lambda and SQS Dead Letter Queues (DLQ) to process images, reducing operational infrastructure costs by 70%.
InterviewPrep: Full Stack Quiz App
A comprehensive learning platform allowing users to master core programming topics through interactive quizzes and curated study materials.
What it does
- •Hyper-detailed, structured learning content across 10 core programming topics (DSA, OOPS, React, etc.).
- •Topic-wise interactive quizzes with randomized and All-in-One testing modes.
- •Modular NestJS backend providing dedicated REST endpoints for learning content and question retrieval.
Peeppa: Price Tracker Engine
A cross-retailer product scraping engine that automatically tracks price drops across major outlets like Amazon, Best Buy, and Target.
What it does
- •Real-time and historic price tracking across automated scrapers spanning diverse e-commerce structures.
- •Dynamic threshold-based email alerts triggering notifications when products drop below target prices.
- •Historical pricing charts mapping price fluctuations dynamically from MongoDB.
2D Interactive Maze Game
A 2D interactive maze game built using Python and Tkinter that generates a new solvable maze every run using Depth-First Search (DFS).
What it does
- •Procedural maze generation using recursive backtracking (DFS).
- •Guaranteed solvable maze from start to end with win detection and smooth frame-based updates (~60 FPS).
- •Grid-based coordinate-to-pixel transformation supporting dynamic scaling.
RawhPlayer: Desktop Media Player
A lightweight desktop media player application focused on custom UI design and core playback functionality.
What it does
- •Audio playback controls (Play / Pause / Stop) with manual file handling and media loading.
- •Modular code structure separating event-driven UI components from core logic.
- •Tight state management across playback transitions (idle → playing → paused).
Facial Expression Recognition with PyTorch
A machine learning project designed to accurately classify human facial expressions from images using a trained PyTorch model.
What it does
- •Preprocessing of facial images for robust feature extraction.
- •Training a convolutional neural network (CNN) for precise expression recognition.
- •Achieving high accuracy in classifying diverse expressions such as happiness, sadness, anger, and surprise.
Image Classifier using Tensorflow
A comprehensive deep learning model building experience producing high validation accuracy by classifying images into organized folders.
What it does
- •Enhanced the dataset with robust data augmentation techniques to drastically improve model accuracy and robustness.
- •Built and trained a convolutional neural network (CNN) with optimized layers and dropout regularization.
- •Developed an automated pipeline to actively classify images and dynamically move them into respective category folders based on predictions.