Built distributed ETL pipelines with Apache Spark (Scala) and Airflow to process large-scale VCF datasets from AWS S3 into Snowflake for analytics and model training.

Integrated real-time clinical decision support via .NET Core Web API backend, React/Next.js frontend, and embedded Tableau dashboards

Leveraged Docker, AWS (S3, EC2), and Snowflake for scalable, containerized deployment and cloud-based healthcare data processing.

Implemented safe, concurrent process handling leveraging Rust's ownership model to eliminate memory safety issues.

Integrated core Unix features: signal handling, environment variable expansion, command history, and configuration support.

Utilized key Rust crates (nix, crossterm, clap, anyhow, serde) for system calls, terminal control, CLI parsing, and robust error handling.

Architected a secure, cloud-native execution pipeline with AWS EKS Kubernetes Jobs, Docker containers, and strict resource/time limits for sandboxed script execution.

Integrated high-performance data processing using Apache Spark (PySpark), Pandas, and NumPy for large-scale backtesting and statistical analysis.

Designed and deployed a scalable backend with Django REST Framework, AWS RDS (PostgreSQL + SQLAlchemy), Amazon S3 for persistent storage, and Nginx for hosting and reverse proxy.

Implemented advanced rendering techniques such as recursive reflections, soft shadows, volumetric atmospheric effects, and procedural material models for photorealistic imagery.

Optimized mathematical computations with SIMD-inspired vector operations and cache-friendly data layouts to accelerate 3D vector math, ray-object intersections, and lighting calculations.

Designed and integrated spatial acceleration structures (BVH, Octree) and memory pooling strategies to minimize allocation overhead and maximize intersection throughput.

Implemented secure passwordless authentication using OTP login and integrated Stripe payments to enable seamless lesson purchases for merchant tutors.

Built AI-powered features leveraging LangChain, Chroma vector DB, and Retrieval-Augmented Generation (RAG) to automatically generate quizzes and transcribe video call audio into actionable student feedback.

Architected caching layers using Apache Kafka, Elasticsearch, and Redis to optimize search, chat log retrieval, and session data performance at scale.

Developed a dynamic formula parser allowing users to input custom fractal formulas (e.g., z^2 + c, conj(z)^3 + c), enabling flexible rendering of diverse fractal types and patterns.

Created an intelligent zoom system with boundary detection and adaptive quality scoring to prevent zoom drift and maintain visual interest during infinite fractal zoom sequences.

Designed efficient tile-based rendering architecture with memory pooling and cache-friendly coordinate calculations to optimize resource usage and accelerate frame generation.

Implemented core CNN components including custom 2D convolution, batch normalization, max pooling, ReLU activation, and Adam optimizer without relying on deep learning frameworks.

Designed and executed data preprocessing and augmentation pipelines to improve model robustness on complex multi-digit street view images.

Optimized CNN architecture with He initialization and im2col convolution technique, resulting in efficient training (~45 minutes for 30 epochs) and fast inference (~15ms/image).

Implemented core GAN components including generator and discriminator networks, binary cross-entropy loss, and backpropagation with custom activation functions (ReLU, Sigmoid).

Designed and synthesized a custom dataset of 28×28 alphabet glyphs with noise and random transformations to train the GAN for diverse letter generation.

Developed an end-to-end adversarial training loop alternating updates of generator and discriminator, demonstrating understanding of GAN convergence dynamics.

Implemented custom logic to differentiate known individuals and strangers, triggering email alerts with face image attachments.

Automated secure email notifications via SMTP SSL to multiple recipients upon recognized arrivals or unknown visitors.

Managed face encoding storage and comparison using JSON serialization with custom NumPy encoder/decoder.

Implemented dynamic location-based services for food banks, shelters, clothing drives, and mental health clinics to support people facing poverty.

Created multiple reusable React components to render interactive maps and location listings based on user geolocation permissions.

Designed accessible UI modals with clear messaging for users without location access, enhancing user experience and engagement.

Implemented dynamic obstacle spawning and smooth player controls for engaging, responsive gameplay.

Integrated coin collection mechanics enabling vehicle and parts upgrades to enhance player progression.

Optimized game performance for mobile devices ensuring smooth frame rates and low latency.