Marc Victor Velasquez

Automation and AI solutions engineer with production experience building n8n workflows, REST/webhook/OAuth integrations, and Python backend services. Shipped 8+ deployed automations connecting CRMs (HubSpot), operations tools (ClickUp), an accounting platform, and data systems, plus LLM-powered and agentic workflows. Comfortable scoping, delivering, documenting, and supporting adoption of solutions for stakeholders end to end.

AI Specialist Intern, AI Engineering & Automation Track

• Architected multi-agent and multi-stage automation workflows for web scraping, research, content classification, data preparation, and validation.

• Built internal AI tooling including Notion custom agents (Article Formatter, SGPB Grabber) and screenshot-capture experiments using n8n, Firecrawl, and ScrapingBee for website/Facebook capture and resizing.

• Delivered an end-to-end, 3-phase outreach automation in n8n: AI-personalized cold outreach, automated reply classification (YES, INQUIRY, REFERRAL, NO, BOUNCED), ClickUp status routing, HubSpot lead/company creation, and AI-generated follow-up interview questions saved to Google Docs.

• Stabilized production automations with webhook fixes, rate-controlled sessions, crash safeguards, and human-in-the-loop Gmail draft review, sending 20+ real outreach emails.

• Built and deployed a 5-step job-prospecting pipeline using Bright Data, OpenRouter, Google Sheets, and Gmail drafts, with deduplication, scheduled runs, production monitoring, evidence-gated contact-email enrichment, and MX validation.

• Engineered Google Search operator/Boolean query strategies for 129 roles and processed 697 unique direct-to-source job listings after filtering and deduplication.

• Built HR automation workflows using TeamLogger, ClickUp, and Google Sheets for check-in verification and idle-time monitoring; fixed critical logic and merge-node failure modes for production rollout.

• Used OpenClaw and Hermes Agent setups as agentic AI research and execution environments during the internship, accelerating automation design, tooling research, and implementation; later presented the setup and configuration to executives for potential company-wide adoption.

• Built a retrieval-augmented generation (RAG) pipeline using LangChain via n8n, embedding documents into a vector store and retrieving relevant context to ground LLM outputs.

• Tech Stack: n8n · LangChain · Bright Data · Firecrawl · ScrapingBee · OpenRouter · LM Studio · HubSpot · ClickUp · TeamLogger · Google Sheets · Airtable · Gmail · WhatsApp · Notion AI Custom Agents · OpenClaw · nanoClaw · Hermes Agent · Claude · Gemini · DeepSeek · Qwen · Mistral

AI Specialist Intern, AI Content Quality & SEO Automation Track

• Generated and published 772 SEO-compliant brand articles to kaizenaire.com using Grok and WordPress, passing premium SEO, readability, and inclusive-language checks; vetted and rejected 293 additional candidates for content, security, and compliance reasons.

• Engineered a single-shot Grok/SuperGrok system prompt that outputs a complete brand article plus Yoast SEO metadata in one pass; eliminated sub-header omissions, unwanted tables, over-length meta descriptions, and background-research preambles.

• Added a session-isolation protocol to stop Grok memory/context leakage between brand generations.

• Built SGP-Clipper, a Chrome extension integrating public business-registry data into production prompts for Grok and Google Sheets, reducing per-company processing time from ~15 minutes to ~7–8 minutes (~40% faster).

• Designed pre-research prompt setup that reduced LLM processing time from ~1m40s to ~1m10s; refined a multi-agent model to ~1m8s and reduced the image-resize step 3× from 45s to 15s.

• Operated a contact-data enrichment run for 15k+ Singapore-company records, validating extraction and preparing clean imports for the lead database.

• Hosted a workshop for company department heads and employees on automation systems developed during the internship period and authored an internal technical implementation guide for lead-database enrichment.

• Tech Stack: Grok (SuperGrok) · Claude · Gemini · DeepSeek · Qwen · Mistral · WordPress · Yoast SEO · SGPBusiness · JavaScript · Chrome Extensions

Full-Stack Web Developer

• Built and delivered client web applications end to end, translating client requirements into working full-stack systems

• Developed and customized WordPress sites using themes, plugins, and page builders; gained hands-on exposure to Shopify, OpenCart, and Duda for site and store builds

• Delivered for international clients under fast-paced, tight-deadline agency-style workflows, optimizing site speed, mobile responsiveness, and code structure for search performance

• Tech Stack: WordPress · Shopify · OpenCart · Duda · Next.js · Express.js · PostgreSQL · Tailwind CSS · JavaScript

Lead AI Engineer | Angeles University Foundation

• Built a dual Random Forest regression ensemble (scikit-learn) predicting flood probability and depth across the Pampanga region, trained on Google Earth Engine-derived flood-frequency and depth-proxy labels (Global Flood Database) and evaluated with R² and feature-importance analysis

• Engineered a 6-feature geospatial dataset (elevation, slope, flow accumulation, road/landslide proximity, population density) sampled over a ~1 km prediction grid via GeoPandas/rasterio, with StandardScaler normalization and an 80/20 train/test split

• Serialized the trained models and feature scaler with joblib (.pkl) and exported a GeoPackage risk map for fully offline integration into the Flutter survival-beacon app

• Constructed a multi-source data pipeline incorporating the OpenMeteo API for real-time meteorological tracking and feature input

• Utilized Google Gemini API to design an intelligent advisory system generating hyper-localized storm preparation checklists informed by the model's flood-risk predictions and live weather telemetry

• Tech Stack: scikit-learn (Random Forest) · GeoPandas · rasterio · joblib · Google Earth Engine · OpenMeteo API · Google Gemini API · Flutter

Full-Stack AI Engineer

• Trained a YOLOv11 stacking ensemble to detect ACL and meniscus tears from MRI coronal plane data (fastMRI and fastMRI+ datasets).

• Achieved 75.17% mAP, 88% FROC, and 73.50% F1-score for bounding-box prediction accuracy; containerized the model and Next.js web application using Docker and AppWrite.

• Presented research at AUF's MAGNUS 2026 and IRCITE 2026.

• Collaborated with AUF CAMP’s faculty on model validation, systems testing, and usability evaluation.

• Tech Stack: YOLOv11 · PyTorch · Next.js · Docker · AppWrite · fastMRI / fastMRI+

BS in Computer Science (2026-06)