about
experience
contact
scorpionlabz
nikhil yates is a versatile full-stack engineer and entrepreneur based in Toronto. With a solid foundation in computer science and passion for building innovative solutions, his interests span Web3, generative AI, data interactivity, robotics, and user experience design.
TLDR: i like stuff(building + designing + learning)
want to collaborate? let's connect.
experience
overview
Designed and productionized LLM-powered systems, turning proofs-of-concept into scalable, fault-tolerant deployments delivering immediate business value. Built domain-specific RAG pipelines (ingestion, recursive chunking, embeddings, vector similarity search, grounded generation) and semantic search / conversational interfaces for both structured and unstructured data. Established LLM testing & evaluation workflows and implemented cost-aware inference strategies for reliable production usage.
notables
- Productionized LLM-powered generative AI systems with fault-tolerant architectures and graceful degradation
- Built domain-specific RAG pipelines to improve accuracy and reduce hallucinations
- Implemented metadata-aware retrieval, similarity tuning, contextual reranking, and prompt grounding
- Established evaluation workflows (structured outputs, confidence scoring, LLMs-as-judge, human-in-the-loop review)
- Reduced AI operating costs by 70–95% via cost-aware inference strategies and semantic drift detection
- Shipped API-first AI microservices in Python and TypeScript using Docker + CI/CD
key projects
Agentic Research Orchestration System
- Architected a deterministic orchestration system using Azure Durable Functions to manage multi-stage AI research workflows, implementing yield-based synchronous flow control and intelligent sub-orchestrator composition.
- Engineered a hybrid multi-LLM architecture that delegates tasks between local Llama 3.1 inference and GPT-4o-mini; integrated a semantic drift detection system (using cosine similarity on 1536-dim vectors) that reduces compute costs by 70-95% on stable trends.
- Built a Next.js chat interface featuring real-time WebSocket polling and dynamic visualization of structured research data, backed by a hybrid storage layer (Supabase for relational metadata, Azure Blob for RAG document storage).
IoT Biomechanics Analysis Platform & Wearable
- Prototyped a custom wearable by integrating multi-axis IMU sensors onto a compression sleeve; wrote low-level protocol handlers to parse quaternion, accelerometer, and gyroscope streams at 100Hz over BLE.
- Developed a real-time visualization pipeline using React Three Fiber and WebSockets, applying sensor fusion algorithms to compute joint angles and correlate acceleration spikes with biomechanical events.
- Implemented a context-aware RAG system using Supabase pgvector and custom reranking to ground live sensor data against medical research, utilizing strategic chunking and ACL-specific boosting for evidence-based analysis.
Social Graph Mobile Architecture & Analytics Engine
- Engineered an offline-first mobile architecture using TanStack Query and Supabase, implementing optimistic updates and a polymorphic UI adapter pattern that reduced code duplication by 60%.
- Designed a serverless event architecture using Supabase Edge Functions (Deno) and Realtime subscriptions to handle high-volume push notifications, badge logic, and geofencing with minimal latency.
- Built a companion Next.js analytics dashboard capable of aggregating complex relational data (user engagement, retention, meetup participation) and generating dynamic PDF reports via server-side API routes.
tech
overview
Built and deployed scalable web applications and AWS infrastructure with robust architectures and CI/CD. Led client-facing delivery through system design, UX improvements, and stakeholder collaboration, contributing to measurable e-commerce growth.
notables
- Built and deployed scalable web apps (React, Node.js, PostgreSQL, AWS)
- Contributed to a 100% increase in e-commerce sales via improved product and UX delivery
- Designed CI/CD pipelines and optimized AWS infrastructure (S3, EC2, Lambda) for cost and reliability
tech
overview
Conducted market research, financial modeling, and KPI analysis to support multi-million-dollar software acquisitions. Built Python tooling to accelerate sourcing and improve data-driven deal execution.
notables
- Built Python-based tooling that reduced SaaS lead identification time by 50%+
- Conducted market research, financial modeling, and KPI analysis supporting multi-million-dollar acquisitions
tech
overview
Worked on the ESSENCE CubeSat. Developed embedded software for a CubeSat onboard computer in a high-reliability environment. Authored technical documentation for CSA-facing review supporting Test Readiness approval.
notables
- Developed embedded software for a CubeSat onboard computer using C and FreeRTOS
- Authored technical documentation presented to the Canadian Space Agency enabling successful Test Readiness approval
tech
overview
Maintained and improved reliability across 20+ systems through diagnostics, hardware upgrades, and NAS deployments.
notables
- Improved reliability across 20+ systems via diagnostics and hardware upgrades
- Deployed NAS solutions and maintained on-prem IT operations
tech
skills
nikhil yates © 2024