Dan Sotnik
Senior AI Engineer @ OrderGrid
About
Senior AI Engineer with 6+ years shipping production AI systems and full-stack SaaS at $200K+ weekly GMV scale. I architect and build AI infrastructure from the ground up — most recently designed and built Corpus, a production MCP tool on AWS that indexes Google Drive, Jira, and codebase artifacts into Qdrant vector DB, forming the foundational agentic AI infrastructure for enterprise automation. Specialized in agentic systems, RAG pipelines, and LLM integration: self-hosted LLM inference in secured AWS VPC for customers with strict data residency requirements, AWS Titan for embeddings, and Google Gemini VLM for computer vision pipelines.Recent impact:- Resolved ~70% of unidentified SKU scans via hybrid vector-embedding matching — identifying misrouted cross-location inventory in real time- Semantic search across 100K+ records at <500ms p99 — powering computer vision pipelines and real-time SKU retrieval- Reduced fulfillment time ~35% across 15K+ weekly deliveries- Eliminated third-party API dependencies for enterprise customers with strict data residency requirementsStack: Python, Node.js, Qdrant, MongoDB Atlas Vector Search, LangChain, LangGraph, AWS, Docker, Gemini, Qwen, MCP.
Canada
Ottawa
Computer Software
Retrieval-Augmented Generation (RAG), Large Language Models (LLM), Model Context Protocol (MCP), LangChain, Self-Hosted LLMs, Node.js, Sentence Transformers, Tool Calling, Python (Programming Language), JavaScript, Vue.js, Qdrant, AI Infrastructure, Continuous Integration and Continuous Delivery (CI/CD), Terraform, Kubernetes, Redis, Microservices, WebSockets, REST APIs
Experience

Senior AI Engineer
Toronto, ON
- Architected multi-stage VLM harness (Gemini) — split-pass recognition, MongoDB Atlas Vector Search matching over 350K-item catalogue, confidence-gated classification, re-scan verification — converting one shelf photo to restock tasks at >80% task precision in <60s. Cut manual audit time 20×; avoided ~$30K/store shelf-robot capex. - Architected and built Corpus, org-wide production RAG platform on AWS (Qdrant + Bedrock, MCP) serving 5 teams; cut prompt tokens 65%, embedding cost 100× via diff-based incremental ingest; p95 ~250ms hybrid retrieval with OAuth 2.1 and fail-closed multi-tenant ACL. - Architected and built Coffre, a Python MCP server (AWS ECS Fargate, Terraform) unifying 4 prod data systems — MongoDB, PostgreSQL, Redis, AWS CloudWatch — behind one governed LLM-agent interface. Read-only by construction with writable dev sandboxes for safe experimentation; 3-layer authz-bypass defense, fail-closed ACLs, OAuth 2.1/SSO, sub-5s p95. Gave agents governed data access, replacing ~20 per-system integrations.

Senior Software Engineer
Toronto, ON
- Architected and built AI-powered inventory identification using vector embeddings (Gemini) in MongoDB Atlas Vector Search with multi-strategy matching (barcode, fuzzy, cosine similarity), resolving ~70% of unidentified SKU scans — identifying misrouted cross-location inventory in real time. - Designed semantic search across 100K+ records at <500ms p99 on MongoDB Atlas Vector Search, powering agentic AI pipelines and Gemini VLM computer vision — cutting manual search time by 70%. - Productionized batch image pipeline processing 750K+ product images via CSV ingestion with streaming I/O, replacing a 2-FTE manual workflow and reducing catalog onboarding from weeks to hours at 99.9% success rate. - Delivered Express Multi-Order Picking with atomic MongoDB transactions and real-time stock validation, reducing fulfillment time ~35% and doubling picker throughput across 15K+ weekly deliveries ($200K+ GMV). - Built Redis-backed presence system with TTL heartbeat tracking, supporting 2,000+ concurrent users across 40+ locations and cutting API latency ~40%. - Shipped Rider Count feature (Node.js, MongoDB, Vue.js, Python) across 42 locations in 4 countries; surfaced 9.3% rider allocation anomaly triggering a cost-saving operational review.

Software Engineer
- Cut paint time 3x on data-heavy components by designing a custom virtualization mechanism powered by chunked rendering for large lists and tables. - Hardened web application security by implementing dual-lifecycle JWT authentication, inactivity-based forced logout, and a role/privilege hierarchy, eliminating known security incidents. - Refactored 200+ APIs using algorithmic improvements and engineering best practices, reducing Lambda compute costs by ~10%. - Owned the IaC layer (Docker, Kubernetes, Terraform), automating ~70% of cloud service deployment. - Cut Unit and E2E test launch time 30% by integrating Inquirer.js interactive commands into the test execution tool. - Drove migration of autotests from Express.js to Nest.js by consulting on architecture and systems design, accelerating the delivery timeline ~20%. - Led code review process improvements based on GitLab Handbook practices, increasing team KPI by 17%. - Collaborated with product and design teams; mentored backend engineers on frontend systems.

Purchasing Department Specialist
• Managed the electrical automotive parts panel with 20+ suppliers, including Multimedia, lighting, HVAC. • Launched Renault Dakar and negotiated with suppliers, achieving an 8% saving from electrical parts turnover. • Contracted special conditions for spare parts production and logistics, increasing availability by 20% and on-time shipments by 60%.
Education

International Relations and Affairs
Additional completed courses: Information Science and Databases, Fundamentals of Mathematical Analysis, Informative Heuristics, Internet System, Project Management, Cross-Cultural Management. The University founded in 1991 on the basis of the Moscow State Institute for History and Archives (1930) is one of the major actors in the Higher Education Reform in the Humanities in Russia. RSUH ranks among the TOP 200 Higher Education Institutions according to QS University Ranking: BRICS 2014 and has expertise in organizing inbound and outbound academic mobility between EU and Russia within EU projects.
Dan Sotnik's Contact Information
Phone
Find the Right Leads
Find Verified Contact Data
What LeadContact does well
Find verified emails, phone numbers, and decision-makers with 98% accuracy.
Find Leads
Find the right people by company, role, industry, location, and more.
925M+ professional profiles

Find Emails
Access verified email addresses for your target contacts.
657M+ emails

Find Phone Numbers
Get cross-validated phone data from multiple top sources.
239M+ phone numbers

More Accurate. Lower Cost.
Find contact data in 1 tool with 98% accuracy
LeadContact integrates leading enrichment tools to deliver more accurate contact data—without paying for each one.
Great conversations start with the right contact.
It’s time to find yours.



