I build n8n pipelines, AI agents, and data systems that actually run in production — connected to BigQuery, Supabase, and Vercel.
Webhook-triggered pipelines that pull data from BigQuery or Supabase, run it through an LLM, and produce structured docs — zero manual steps.
Multi-source enrichment pipelines that cross-reference patent DBs, academic records, LinkedIn — then score and route leads automatically.
AI-powered onboarding agents, pairing bots, and digest engines for Telegram communities — fully automated with Supabase-backed state.
Data extraction, transformation, and loading between any stack — APIs, databases, spreadsheets. Clean, validated, structured output every run.
Vector-search pipelines using Pinecone or pgvector — retrieve the right context, combine with live data, generate accurate structured outputs.
Multi-step AI agents that plan, retrieve, execute, and loop — built in n8n with proper error handling, retry logic, and structured outputs.
Discovery
Understand the workflow, existing stack, and where the manual pain actually is.
Architecture
Map out nodes, data sources, error paths, and outputs before touching n8n.
Build
Wire it up — n8n, Supabase, BigQuery, Vercel. Real integrations, real data.
Test & Refine
Edge cases, error handling, retry logic. Only prod-stable pipelines get signed off.
Shipped
Handoff with docs, monitoring, and a walkthrough — so your team can own it.
Rebuilt a mock-data dashboard template into a real DeFi analytics product — live subgraph data, IL calculator, address lookup, and full TS + test coverage.
End-to-end GoHighLevel automation: lead capture through retention and win-back, with AI-enhanced nurture and behavior-based branching.
Multi-source enrichment cross-referencing patent DBs, LinkedIn, and academic records via n8n + Supabase.
Intelligent RAG agent combining vector search and live BigQuery queries to generate structured document outputs.
AI-powered Telegram onboarding agent and Random Coffee pairing engine — Supabase-backed state, fully automated.
3-workflow n8n pipeline: weekly data ingestion from GitHub + public API, new-events discovery with team email digest, and on-demand Telegram bot delivery with paginated CFP results.
A retrieval-augmented document generation pipeline for a UK startup client: vector search over prior examples combined with live structured data, automatic document generation, and reviewer sign-off built into the flow. Tone and structure adapt per recipient to avoid a templated feel.
Submit a name and LinkedIn URL; the system scrapes public profile data and scores the candidate against a fully editable, multi-criteria fit rubric via an admin panel. Verified end-to-end against real candidates.
Weekly content digest with a searchable, filterable archive and CSV export, plus email and Telegram subscription delivery. A companion matching layer scores relevance against uploaded client profiles by extracting tags from PDFs via AI.
A structured NDA and services-agreement template system matching WanderingTech's actual fixed-price, 50% deposit engagement terms — ready to fill in and send for any new client.
An independent exploration of multi-criteria candidate scoring, built from scratch with its own logic and synthetic sample data — not derived from any employer's codebase. Enter a name and target domain, get a structured fit score across four rubric dimensions, plus a sortable, filterable candidate pipeline view.
A concept demo for a Meta-Ads-to-close lead management system for trade businesses — built independently, with fully synthetic sample data. Dashboard with conversion/pipeline-value stats, a drag-free Kanban pipeline (New → Contacted → Qualified → Quoted → Follow-up → Won/Lost), and a lead detail view with a full activity timeline and an automation on/off toggle for human takeover.
A two-sided marketplace connecting travelers with local storage partners across Cagayan de Oro's transit hubs. Six-step booking flow with transparent, itemized pricing and an optional insurance add-on; pickup verified by matching a scanned QR code against the tamper-seal number logged at drop-off. Separate consoles for partner counters (daily drop-off/pickup board, payouts) and marketplace ops (partner approval, dispute resolution).
An end-to-end invoicing pipeline replacing manual, spreadsheet-based billing — automated generation, status tracking, and reminders. In active development; full case study coming once it ships.
Tools & Tech Stack
I build automation systems that actually run in production.
Currently Technical Project Manager at Immigram (UK), shipping n8n pipelines connecting APIs, databases, and AI models — from lead enrichment at scale to intelligent agentic workflows. Backends are BigQuery and Supabase. Deployments on Vercel.
Background in mechanical engineering and data analysis — financial tracking pipelines, multi-source datasets, and dashboards at ReCool Technologies. That data ops foundation makes the automation work cleaner.
One year deep into this. Still learning fast, still shipping. If your team is sitting on manual processes that should've been automated yesterday — that's the kind of problem I enjoy.
What I Automate
Experience
Technical PM — AI & Automation
Immigram · UK
Freelance Data Analyst
Independent
Research Analyst
ReCool Technologies
Currently Building
WanderingTech CRM
A personal lead pipeline dashboard backed by Supabase — built to manage freelance projects end-to-end.
GoHighLevel — Level 3 Certified
Extendly
Google Project Management Professional
Coursera
Google Data Analytics Professional
Coursera
Data Analysis in Tableau
DataCamp
Automation & Integration
Data & Infrastructure
Dashboards & BI
"Jam was hands-on with our AI-powered onboarding bots, prospecting pipelines, and intelligence dashboards. His work helped provide the data we needed to make better decisions and grow the business. He's reliable, communicates clearly, and genuinely enjoys the nuts and bolts of how automations work. I believe he has a very bright future ahead."
Working together? I'd love to feature your experience here.
Book a Call
Schedule a Session →
Location
Philippines · Remote
Availability
Open to Projects
A six-page Uniswap V3 analytics product pulling real on-chain pool, swap, and LP position data from The Graph — rebuilt from a mock-data template into a working data product with its own quant tooling.
Started from a template where every dashboard — pool performance, LP analytics, live trading, risk management — rendered hardcoded mock arrays. No real data, no working exports, fake "live" updates via setInterval.
SolutionRebuilt the data layer on The Graph's Uniswap V3 subgraph and CoinGecko behind React Query hooks, with an automatic demo-mode fallback so the app still runs fully without an API key. Migrated the whole codebase to TypeScript, and added an impermanent-loss calculator with the real concentrated-liquidity V3 math, a read-only address lookup for any wallet's LP positions, working pool comparison/watchlist/CSV export, and dark/light theming — plus Vitest coverage and CI.
ImpactEnd-to-end GoHighLevel automation architecture: lead capture through retention and win-back, with behavior-based branching, AI-enhanced nurture sequences, and a reactivation loop feeding back into Stage 2.
Client had a functional GHL account but every stage of the customer journey required manual intervention — leads weren't being nurtured, post-call follow-ups were inconsistent, and churned clients had no re-engagement system.
SolutionDesigned and built a 7-stage automated lifecycle — from landing page capture to reactivation — with conditional branching based on engagement, lead scoring thresholds, outcome-based post-call flows, and a lapsed-client win-back loop that re-enters at Stage 2.
ImpactA Vercel-deployed internal tool that launches and configures six n8n discovery pipelines to surface high-value technical candidates by cross-referencing global patent databases and academic records.
Manually searching for technical talent across patent filings, Google Scholar, and LinkedIn was taking hours per candidate with no standardized scoring — impossible to scale without a larger team.
SolutionBuilt a configurable pipeline launcher where the operator selects a discovery strategy, sets scoring thresholds, picks target countries from a tiered 37-country list, and fires a structured payload to the appropriate n8n webhook.
ImpactA full-stack freelance pipeline tool — from automation scoping form to real-time Kanban board backed by Supabase.
No structured way to track inbound leads, understand what they needed before a call, or move projects through stages without relying on spreadsheets and memory.
SolutionBuilt a portfolio site with an embedded 6-step automation scoping form. On submit, data inserts directly to Supabase and the lead appears live in a password-gated CRM — complete with Kanban board, full brief modal, and Calendly pre-filled with the client's details.
ImpactAn independent, self-contained demo exploring multi-criteria candidate scoring — no backend, no real data, no scraping.
Wanted to demonstrate a talent-scoring UX pattern — structured rubric breakdown, verdict, sortable pipeline — without touching any employer's real system or data.
SolutionBuilt a fully client-side app from scratch: a deterministic scoring function (seeded from name + domain, no network calls) drives a rubric breakdown across four dimensions, plus a sortable/filterable candidate pipeline view seeded with fabricated sample data.
ImpactA concept demo for a Meta-Ads-to-close lead management system for trade businesses — no backend, no real leads, fully synthetic sample data.
Wanted to demonstrate a full lead-to-close CRM pattern — Kanban pipeline, activity logging, human-takeover control — for a client pitch, without building on any employer's real system or client data.
SolutionBuilt a fully client-side app from scratch: a seeded sample lead set drives a 6-stage Kanban board, a searchable/sortable leads table, and a lead detail drawer with a generated activity timeline (lead received → confirmation → sequence → reply detected → quote → won/lost) and an automation on/off toggle representing human takeover.
ImpactAn end-to-end AI pipeline that retrieves context from a vector database, queries live BigQuery data, and generates structured outputs — deployed on Vercel, triggered via n8n.
Complex document workflows required manually retrieving information from multiple sources and drafting structured outputs — taking 2–4 hours per cycle with inconsistent results.
SolutionBuilt a RAG pipeline in n8n using Supabase (pgvector) as the vector staging layer and BigQuery as the production data backend.
ImpactTwo Telegram bots: an AI-powered onboarding agent and a Random Coffee pairing engine.
Onboarding was manual and inconsistent. Weekly team connections (Random Coffee) had no follow-up system.
SolutionBuilt two bots. The onboarding agent uses a conversational AI persona to guide new members and verify profile status against Supabase. The pairing engine runs weekly, generates pairings, sends introductions, and follows up automatically.
ImpactA linked set of n8n workflows that track developer conferences and CFP deadlines — ingesting from external sources, deduplicating, refreshing stale data automatically, and delivering results to Telegram on demand.
Stale and past events were showing up in digests because filtering had no lower date bound, and previously-missing CFP details never got backfilled once a record was written.
Root Cause FixedRebuilt the date-filtering logic with a proper two-sided window, and switched the write strategy from append-only to an upsert pattern so records self-heal as new data becomes available.
ImpactA structured funding intelligence pipeline that tracks, scores, and prioritizes capital programs — replacing manual grant research with automated scored shortlists.
Researching funding opportunities was entirely manual — no consistent scoring, no prioritization system, and no way to track status across multiple programs.
SolutionBuilt an n8n pipeline that ingests opportunities, scores each across 12+ vetting metrics, and outputs a ranked structured shortlist covering programs up to £5M.
ImpactA Google Sheets KPI system benchmarking 22 global talent communities across 7 platforms.
No structured view of how target communities performed across platforms — impossible to prioritize partnerships based on actual engagement data.
SolutionBuilt a multi-tab Google Sheets dashboard aggregating follower counts, engagement rates, and platform dominance metrics across 22 communities.
ImpactGoHighLevel — Level 3 Certified
Extendly · 2025
Google Project Management Professional
Coursera · 2025
Google Data Analytics Professional
Coursera · 2025
Lean Six Sigma Yellow Belt
MF Business School · 2025
Lean Six Sigma White Belt
MF Business School · 2025
Data Analysis in Tableau
DataCamp · 2024
Registered Mechanical Engineer (RME)
PRC Philippines
Automation & Integration
Data & Infrastructure
AI & Agentic Systems
Dashboards & BI
Project & Operations
Before we talk, tell me a bit about your situation. This takes 3 minutes and means our call goes straight to solutions — no back-and-forth on basics.
Brief received — now pick a time.
I'll come to the call reviewed and ready with a rough plan.