Relevant Coursework: Data Structures and Algorithms, Machine Organization, Systems Programming

• Built a cross-platform AI automation framework that enables users to schedule, chain, and run autonomous agents across OS-level (FastAPI + Claude Computer-use) and browser-level (Playwright + Docker) environments with persistent memory
• Architected a distributed backend using Redis, QStash, and SQL, enabling stateful memory of agent via LangGraph
• Designed real-time infrastructure via WebSocket + Redis pub/sub, shipping the platform to 230+ beta users within 4 weeks

• Built a serverless NestJS backend with AWS Lambda, API Gateway, DynamoDB, enabling scalable agentic workflows
• Deployed a distributed task orchestration system with Redis and BullMQ, supporting concurrent job execution of agents
• Built CI/CD pipelines with GitHub Actions and Terraform, leveraging Docker to streamline deployments to 200+ users

• Developed an orchestrator system delegating user queries to concurrent AI agents, simulating distributed decision-making
• Integrated DALL-E for image generation in AI-powered hint features, transforming educational content into interactive activities

• Implemented Infrastructure as Code (IaC) practices with Terraform to manage AWS resources, including EC2, S3, and RDS
• Monitored real-time system performance using Datadog and AWS CloudWatch, applying alert-driven DevOps workflows
• Provisioned a self-hosted Jenkins cluster with autoscaling EC2 agents and S3 artifact storage, reducing build times by 40%
• Collaborated in an Agile team environment using Git for version control, driving iterative development and CI/CD practices

• Implemented APIs and UIs for a no-code Workflow Automation platform leveraging Nextjs, Fastapi, and PostgresQL
• Designed integrations with OpenAI, Slack, Discord, Notion, Airtable, and Google Drive to automate notifications, custom messaging, and database entry creation, enhancing cross-platform connectivity, reducing manual human effort by 80%

• Built an SQL-like database in C with custom heap allocator, memory-mapped file storage, and page-based virtual memory
• Designed and implemented B-Tree indexing to enable logarithmic search complexity and fast lookups across file-backed storage
• Engineered a file I/O system using `mmap`, `open`, and `write` syscalls, manually serializing/deserializing row data
• Created a custom heap allocator with pointer arithmetic, record slot management, and dynamic memory reuse across pages

• Implemented limit order book supporting FIFO/pro‑rata matching in C++ using custom price-level queues and hash maps
• Designed structures enabling near‑constant lookup/insertion; profiled with gperftools, achieving ~1M orders/sec
• Simulated HFT scenarios including spoofing and iceberg orders; generated time-series features for downstream analytics

• Built a vectorized backtester over 10+ years of equity data using pandas/NumPy, with dynamic rebalancing and position sizing
• Integrated WebSocket price feeds from Polygon and Alpaca, added transaction‑cost modeling to simulate production conditions
• Employed Bayesian hyperparameter tuning with Optuna to optimize Sharpe ratio across differing market regimes

• Built a data aggregator to ingest market tick data from IEX and Polygon, handling over 1M ticks/sec using Kafka pipelines
• Instrumented with Prometheus + Grafana, tracking ingestion rate, latency percentiles, and cache hit ratios under simulated load
• Used Redis to implement low-latency caching for tick snapshots and last-trade data, exposed via WebSocket and HTTP APIs

Contributed to Alto, a fully automated, agent-driven QA and debugging platform built with Fetch.ai, FastAPI, and PostgreSQL Implemented a modular multi-agent architecture leveraging Python, OpenAI, and CrewAI to orchestrate testing agents Integrated with Instabug for real-time bug reports, GitHub for code retrieval, and Slack/Jira for notification/kanban agents

Fine-tuned ResNet18 for image recognition to match sketches with found images, achieving 76% accuracy in object detection Designed a RAG system to power a chatbot, using SQL databases for knowledge storage and AI models for response generation Developed an item-matching algorithm using PyTorch, SpaCy, and Scikit-learn, improving description-based item retrieval Built a full-stack application with React and MySQL, integrating Prisma ORM for schema management and Clerk for auth

Built a real-time transaction system with Plaid for banking integration, PostgresQL for data storage, and Clerk for auth Developed an AI assistant interface enabling users to query financial data and receive actionable insights in natural language Implemented interactive dashboards with React, Chart.js, and Three.js to visualize spending patterns and financial breakdowns

Built real-time CV pipeline with OpenCV, MediaPipe, DeepFace for face detection and focus tracking via webcam Fine-tuned models in TensorFlow/Keras and PyTorch for attention classification, improving accuracy across pose variations Applied CNN embeddings and transfer learning for robust face representation; integrated Gemini for orchestration

Built a multi-agent system using Fetch.ai uAgents and CrewAI to autonomously launch businesses from a single idea prompt Integrated Shopify, Stripe, Google Ads, and OpenAI SDK APIs to automate store creation, checkout, marketing, and analytics Designed a Flask + Supabase backend and Next.js dashboard, visualizing agent orchestration and workflow states in real time

Built an AI phone agent that autonomously handles real-time calls, navigates IVRs, converses naturally, and reports results Engineered a fault-tolerant speech pipeline with VAPI telephony, ElevenLabs TTS, and Deepgram STT for reliable call flow Developed a React + Django dashboard with SQL backend to manage call logs, transcripts, recordings, and analytics

YC AI Startup School