Projects

Systems I designed, shipped, and measured

Four builds in depth, then the rest of the archive. Every one links to its repository.

AI Co-PilotSaaS

AI Co-Pilot for SaaS Analytics Platform

Built an AI co-pilot enabling natural-language access to a SaaS analytics dashboard via FastAPI and Anthropic function calling, mapping user queries to structured backend calls.
Implemented SSE token streaming and JWT-authenticated services, achieving p95 first-token latency under 2 seconds.
Added prompt-injection detection, role-based access control, and query validation at the request boundary.

<2s first-token latency
SSE streaming
Role-based access

ReactFastAPIAnthropic APISSE

View on GitHub

Agentic AILLM Systems

Multi-Agent Research Assistant

Built a multi-agent research platform orchestrating planner, researcher, critic, and synthesis agents with tool-calling and shared state management to automate multi-step research workflows.
Deployed agent microservices on Kubernetes with inter-agent communication latency below 200ms.

Multi-agent workflows
<200ms latency
Kubernetes microservices

LangGraphClaude APIRedisDocker

View on GitHub

LLM SystemsEvaluation

Agent Arena

Engineered an evaluation system for LLM models providing a universal arena to determine model selection based on accuracy, cost, and latency, alongside a multi-agent study framework. Features offline evaluation, local model support, and deterministic mock models.

Offline evaluation
Cost & latency modeling
Multi-agent framework

PythonAnthropic APIOpenAI APILLM Evaluation

View on GitHub

Multimodal AIClinical AI

AffectiSense

Engineered a multimodal mental health assessment platform fusing neurophysiological (EEG), vocal biomarker (audio), and behavioral (facial expressions) signals for depression screening. Features Modality Dropout for resilience and interpretability mapped to DSM-5 symptom clusters.

Modality Resilience
HIPAA Aware
DSM-5 Mapping

PyTorchFastAPIHuBERTMediaPipe

View on GitHub

More projects

Research prototypes, coursework built past the brief, and earlier production work.

IndusAI

Engineered an industrial AI assistant for PLC fault log analysis; migrated to Google Cloud (Gemini Flash, Qdrant, Cloud Run, Firebase)

100+ environments · >99% redaction · 30x latency reduction

GCPGemini FlashQdrant CloudCloud Run

GitHub repo

Multimodal Depression Analysis

Built a real-time system recognizing depression by extracting features from facial expressions, EEG signals, and speech modules using multimodal deep learning. Supported by peer-reviewed research publications

3 modalities · 91% precision · IEEE published

NLPFeature EngineeringSignal ProcessingPyTorch

GitHub repo

Object Detection & Localization

Trained YOLOv3 on the COCO dataset for real-time object detection and localization on road scenes; related to autonomous vehicle patents and deployed in a braking system prototype

Real-time detection · 2 patents filed · YOLOv3 + COCO

TensorFlowPyTorchOpenCVCUDA

GitHub repo

Stock Index & GDP Forecasting

Developed an ARIMA model built with MATLAB's Econometrics Toolbox to forecast GDP and Stock Indexes of 10 countries. Includes Python-based validation, residual diagnostics, and confidence intervals for economic trend analysis

10 countries · GDP + Stock Index · ARIMA modeling

PythonMatlabARIMATime Series

GitHub repo

Business Location Recommendation System

Architected a data analysis and ML pipeline to find optimal business locations using Wikipedia data and Foursquare API for Toronto neighborhoods; features Folium mapping, K-Means clustering, and geospatial visualization

Toronto neighborhoods · Geospatial clustering · Foursquare API

PythonK-MeansFoliumFoursquare API

GitHub repo