status: open_to_work

Anslem Ebiega

AI Automation & Full Stack Engineer

I build AI systems that do the work — and the platforms they live in.

Toronto, Canada

GitHub
0+years shipping production software
0+products from zero to launch
0+users onboarded & verified
0+countries served
0industries covered by AI assessments

Software that does the work

I'm a full-stack engineer who specializes in making software do the work. I've spent 10+ years shipping production software — and the last several building platforms where AI automation isn't a feature bolted on at the end, it's the engine: a voice agent that answers a business phone line 24/7 and books the appointment while the caller is still on the call, hiring assessments that score candidates across 23 industries with objective, comparable results, a discovery engine that scans 8 platforms and validates startup ideas in minutes instead of weeks, a marketplace where automation deliverables are verified by the work itself rather than by self-reported skills, and annotation pipelines that feed production ML systems with human-in-the-loop quality control.

I hold a Master's in Computer Science (AI) from the University of Jos and a Post-Graduate Certificate in Blockchain & Back-End Development from York University. I've led cross-functional teams, onboarded 12,000+ verified users across 42+ countries, and taken more than seventeen products from first commit to production. My home turf is the seam where LLM pipelines meet real infrastructure — the interface, the services behind it, and the data layer underneath.

AI Automation Systems

LLM-powered pipelines that replace manual workflows end to end — voice agents that answer a live call and book the appointment on it, assessment engines with behavioral scoring, multi-platform discovery crawlers, and human-in-the-loop annotation flows.

Claude APILLM PipelinesAgentic WorkflowsHuman-in-the-Loop

Full Stack Platforms

Production web and mobile products with real-time cores — marketplaces, hiring platforms, collaboration tools. React and Next.js frontends, Node.js and Python services, WebSocket infrastructure, and payments.

ReactNext.jsNode.jsReact NativeWebSockets

Data & Intelligence

The layer that makes automation trustworthy: annotation pipelines with consensus scoring, analytics dashboards, aggregation-driven recommendation engines, and API-ready dataset delivery into ML pipelines.

PythonPandasTensorFlowPostgreSQLMongoDB

$ every system follows the same loop — manual work in, finished work out

The build log

Systems I've architected and shipped — the problem, the machine I built for it, and what changed.

~/builds/rhantaLIVE
01

Rhanta

The business phone line that answers itself

Voice AIAI AgentsTelephonyCalendar BookingOutbound Campaigns
[ problem ]

A missed call is not a missed message, it is a lost customer. Small businesses cannot staff a phone around the clock, so calls that arrive after hours or while the line is busy fall to voicemail — and the caller simply dials the next business on their list. The revenue leaves before anyone knows the call happened.

[ the system ]

I built Rhanta to answer the business line 24/7 and carry a call all the way to a booked appointment rather than to a message someone has to action later. It picks up every call, captures the caller as a lead, and writes real appointments straight into the business's calendar while the caller is still on the phone. The same platform runs the outbound direction too — dialling the existing list back for reminders, offers, and follow-ups — so inbound capture and outbound win-back happen on one system instead of two.

[ under the hood ]
Always on
The line is answered rather than routed to voicemail, on the assumption that a caller who reaches a recording dials the next business on their list — the value is in picking up, not in taking a message.
Booked, not messaged
A call ends with an appointment written into the calendar rather than a note for someone to follow up on, which is the step where a captured lead usually turns back into a lost one.
Both directions
The same platform places outbound calls against the existing customer list for reminders, offers, and follow-ups, so winning back a quiet list and catching new callers are one system rather than two.
[ impact ]
  • Answers the business line 24/7, so after-hours and overflow calls convert instead of reaching voicemail
  • Books real appointments directly into the business calendar during the call
  • Captures every caller as a lead without anyone taking notes by hand
  • Outbound reminders, offers, and follow-up campaigns run from the same platform
~/builds/reworkLIVE
02

Rework Digital

The platform for AI & automation work

ReactNode.jsTypeScriptFirebasePythonProof-of-Work Verification
[ problem ]

Companies want automation — chatbots, AI agents, workflows, data pipelines — but hiring for it is guesswork. Anyone can claim to be an automation expert, portfolios can't be verified, and paying upfront for work that may never ship burns budgets. The AI economy had no infrastructure for trust.

[ the system ]

I architected Rework Digital as a verified marketplace purpose-built for the AI and automation economy. Builders prove their skills through real deliverables — chatbots, workflows, pipelines, integrations, and agents verified with cryptographic proof of work, not self-reported resumes. Companies post a brief for free, get matched with vetted specialists across six categories (AI chatbots & agents, workflow automation, AI integration & APIs, data pipelines & analytics, RPA & document processing, and marketing automation), and pay through milestone escrow: funds sit protected and release only when the shipped work is approved.

[ under the hood ]
Trust
A builder's track record is derived from shipped deliverables verified by cryptographic proof of work, not from a self-reported skills list — the profile is an output of the work rather than a claim about it.
Payments
Milestone escrow holds client funds until the shipped work is approved, so neither side carries the full counterparty risk of an upfront transfer.
Matching
Briefs route across six automation categories — chatbots and agents, workflow automation, AI integration and APIs, data pipelines and analytics, RPA and document processing, marketing automation — with timezone, language, and budget as first-class filters.
[ impact ]
  • 7,000+ verified automation specialists onboarded across 42+ countries
  • Every deliverable verified through cryptographic proof of work — chatbots, workflows, pipelines, and agents
  • Milestone escrow and transparent pricing protecting both clients and builders end to end
  • Six automation service categories with timezone, language, and budget coverage for any team
~/builds/switchassessLIVE
03

SwitchAssess

The hiring evaluation pipeline, fully automated

ReactNode.jsTypeScriptFirebasePythonBehavioral Scoring
[ problem ]

Resumes and unstructured interviews are poor predictors of job performance, and manual screening doesn't scale — companies burn thousands on bad hires because every step of evaluation depends on subjective human judgment.

[ the system ]

I designed and built an AI assessment engine that automates candidate evaluation end to end: career-specific simulations across 23 industries, live coding sandboxes for technical roles, anti-cheating proctoring with real-time behavioral monitoring, and an analysis layer that aggregates Performance (40%), Behavioral Signals (25%), Role-Fit Intelligence (25%), and Integrity Verification (10%) into a single Hiring Confidence Score. Employers assemble multi-stage assessments in a step-by-step builder; the system handles proctoring, scoring, risk-flagging via a Regret Minimizer Report, top-performer similarity matching, and explainable AI-powered candidate comparison — replacing gut-feel screening with comparable, explainable scores.

[ under the hood ]
Scoring
One Hiring Confidence Score composed from four explicitly weighted signals — Performance 40%, Behavioral Signals 25%, Role-Fit Intelligence 25%, Integrity Verification 10% — so two candidates assessed for different roles remain comparable on the same scale.
Integrity
Proctoring runs as live behavioural monitoring during the session and enters the composite as its own weighted term, rather than acting as a pass/fail gate that throws away a candidate on a single signal.
Explainability
Comparisons state why one candidate outranks another, and a Regret Minimizer Report surfaces the risk in a hire before it is made — the scores are arguable, which is the point.
[ impact ]
  • Career-specific assessment simulations deployed across 23 industries
  • Anti-cheating proctoring system with real-time behavioral monitoring
  • Live coding sandboxes with multi-language support for technical evaluations
  • AI behavioral analysis producing composite Hiring Confidence Scores reducing subjective bias
~/builds/finprobLIVE
04

Finprob

Market research on autopilot

ReactTypeScriptPythonSentiment AnalysisFirebaseNode.js
[ problem ]

Validating a startup idea takes weeks of manual research, and the output is usually confirmation bias. Founders didn't need another survey tool — they needed the discovery process itself automated.

[ the system ]

I built an AI discovery engine that scans 8 platforms simultaneously — Reddit, Twitter/X, Product Hunt, Hacker News, G2, Trustpilot, app store reviews, and Stack Overflow — and runs 7 automated analysis methods over the raw conversations: sentiment analysis, pain-point clustering, demand-signal detection, trend velocity tracking, competitor gap analysis, market sizing estimation, and opportunity scoring. Every surfaced idea ships with a validation score, TAM/SAM/SOM market sizing, and a competitor landscape breakdown — no analyst required.

[ under the hood ]
Ingestion
Eight sources are scanned in parallel — Reddit, Twitter/X, Product Hunt, Hacker News, G2, Trustpilot, app store reviews, and Stack Overflow — so a demand signal can be corroborated across communities instead of resting on one loud thread.
Analysis
Seven methods run over the same raw conversations: sentiment analysis, pain-point clustering, demand-signal detection, trend velocity tracking, competitor gap analysis, market sizing, and opportunity scoring.
Output
Every surfaced idea carries a validation score, TAM/SAM/SOM sizing, and a competitor landscape — a founder can inspect the reasoning rather than being handed a verdict.
[ impact ]
  • 8 platforms scanned simultaneously for real-time opportunity discovery
  • 7 AI-powered discovery methods including sentiment analysis, pain-point clustering, and gap detection
  • Automated validation scoring with market sizing and competitor landscape analysis
  • Reduced idea validation from weeks of manual research to minutes of AI-driven discovery
~/builds/loevechLIVE
05

Loevech

Production-grade training data, humans exactly where they belong

ReactTypeScriptPythonNode.jsFirebaseHuman-in-the-Loop QA
[ problem ]

AI teams lose more time fixing bad labels than training models. Annotation vendors are slow, inconsistent, and blind to quality — the opposite of what a production ML pipeline needs.

[ the system ]

I built an enterprise annotation platform designed as an automation pipeline with human checkpoints: automated pre-labeling with human-in-the-loop verification, multi-tier quality assurance with annotator consensus scoring, custom taxonomy and labeling schema builders, real-time quality dashboards, and a RESTful API that delivers quality-controlled datasets straight into existing ML pipelines.

[ under the hood ]
Pipeline
Automated pre-labeling proposes and humans verify: the model absorbs the volume, and reviewer attention is spent only where judgment actually changes the label.
Quality
Multi-tier review with annotator consensus scoring, surfaced on live quality dashboards — disagreement is caught while a batch is in flight rather than discovered by the customer after delivery.
Delivery
A REST API hands quality-controlled datasets straight into existing ML pipelines, with taxonomies and labeling schemas defined per customer instead of forced into a fixed ontology.
[ impact ]
  • Enterprise-scale annotation services spanning customer segmentation to AI safety audits
  • Quality-controlled workflows with multi-tier review and annotator consensus scoring
  • API-ready delivery enabling seamless integration into existing ML pipelines
  • Custom taxonomy and labeling schema support for diverse enterprise use cases
~/builds/colsphereSHIPPED
06

Colz (Colsphere)

Reimagining music industry collaboration

ReactReact NativeNode.jsMongoDBAWSWebSockets
[ problem ]

Independent artists are locked out of the music industry's inner circle. They can upload tracks, but there's no infrastructure for real collaboration — finding a producer, getting heard by an A&R rep, or landing paid gigs. Labels scout through social media and word-of-mouth, missing talent entirely.

[ the system ]

I built Colsphere as a three-sided platform connecting artists, labels, and fans through active participation rather than passive distribution. The web platform (React) and cross-platform mobile app (React Native) share a real-time backend powered by WebSockets. Artists create verified profiles with EPKs that auto-sync from Spotify and socials, post collaboration projects, and build a network. Labels get data-driven A&R tools — deep search with genre, location, and engagement filters, a signing suite for managing demos, and an asset vault for secure stem sharing. Fans enter the 'Sphere' to discover trending artists, vote on demos, and earn member-only perks.

[ under the hood ]
Real-time core
The React web app and the React Native mobile app sit on one shared WebSocket backend, so the project feed, messaging, and opportunity marketplace behave identically on both without a second implementation to keep in sync.
Profiles
Artist EPKs populate themselves from Spotify and social platforms rather than asking musicians to maintain a press kit by hand — the profile stays current because nobody has to update it.
A&R tooling
Labels get deep search filtered on genre, location, and engagement, a signing suite for managing demos, and an encrypted vault so stems and masters can be shared without leaving unreleased work exposed.
[ impact ]
  • Full web platform and cross-platform mobile app shipped to production
  • Real-time collaboration features: project feed, messaging, opportunity marketplace
  • Data-driven A&R toolkit for labels: deep search, signing suite, asset vault
  • Three-sided marketplace connecting artists, labels, and fans in one ecosystem
~/builds/bluemoonSHIPPED
07

BlueMoon

Multi-vendor fashion marketplace for Africa

ReactReact NativeNode.jsMongoDBPaystackAWS
[ problem ]

Africa's fashion industry is fragmented across thousands of independent designers and tailors with no unified digital marketplace. Customers struggle to discover local talent or book services reliably, while vendors lack the tools to scale beyond word-of-mouth referrals.

[ the system ]

I led the development of a full-scale multi-vendor marketplace connecting designers, tailors, and customers on one platform: geo-location-based service discovery, vendor storefronts with product catalogs, integrated booking and payment systems, a vendor management portal, and real-time chat between buyers and sellers. I also established the brand identity and designed the end-to-end UI/UX across web and mobile.

[ under the hood ]
Discovery
Service discovery is geo-located, because the product is tailoring and fitting — the useful result is a vendor a customer can physically reach, not the best-rated vendor nationally.
Vendor side
Storefronts, product catalogues, booking, and a management portal, so a tailor operating on word-of-mouth has somewhere to run the business rather than just a listing.
Payments and chat
Paystack handles transactions and real-time buyer-seller chat carries the fitting conversation, keeping the negotiation that fashion commissions require inside the platform.
[ impact ]
  • 400+ vendors onboarded within the first month of beta launch
  • Full marketplace with geo-location service finder, booking, and payment systems
  • Vendor portal with storefront management, product catalog, and real-time chat
  • End-to-end brand identity and UI/UX designed across web and mobile platforms

Commit history

Eight years of shipping — from banking systems and security automation to AI platforms.

89a13a4Dec 2024 - PresentOntario, Canada

Full Stack Engineer

Rework

  • Architected Rework Digital, the platform for AI & automation work — companies hire verified experts to build AI chatbots, agents, workflows, integrations, and data pipelines, with every deliverable proven through real work instead of self-reported skills
  • Engineered the platform's trust layer: cryptographic proof-of-work verification for automation deliverables and milestone escrow that releases payment only when shipped work is approved
  • Built marketplace flows spanning six automation categories — AI chatbots & agents, workflow automation, AI integration & APIs, data pipelines & analytics, RPA & document processing, and marketing automation
  • Onboarded 7,000+ vetted automation specialists across 42+ countries and led cross-functional sprints from planning through delivery
ddde85dJun 2025 - Jan 2026Toronto

Founder & Full Stack Engineer · Side Project

Colz (Colsphere)

  • Founded and built Colsphere, a three-sided music collaboration platform connecting independent artists, record labels, and fans with real-time project tools
  • Developed the full-stack web app (React, Node.js, MongoDB) and cross-platform mobile app (React Native) on a shared WebSocket-powered real-time backend
  • Automated artist EPKs with profile data that syncs from Spotify and social platforms, plus a discovery feed, opportunity marketplace, and in-app messaging
  • Built data-driven A&R tooling for labels — deep search with engagement filters, a signing suite, and an encrypted asset vault for secure stem and master sharing
dd9b34eApr 2024 - Apr 2025Ontario

Full Stack Engineer (Contract)

BlueMoon

  • Led a cross-functional team building a full-scale multi-vendor fashion marketplace connecting designers, tailors, and customers
  • Engineered geo-location service discovery, vendor storefronts, integrated booking and payment systems, a vendor portal, and real-time buyer-seller chat
  • Established the brand identity and designed end-to-end UI/UX across web and mobile touchpoints
  • Drove early traction with 400+ vendors onboarded within the first month of beta launch
7e2d1b9Jun 2021 - Mar 2023Abuja

Technical Support

United Bank for Africa

  • Provided hands-on technical support for core banking applications serving millions of customers, keeping systems available around the clock within SLA targets
  • Automated recurring reporting pipelines with SQL and Python, eliminating hours of manual work each week
  • Acted as the liaison between business stakeholders and engineering during incident resolution
  • Developed technical documentation and delivered end-user training sessions
fdf45b3Aug 2020 - PresentRemote & In-Person

Coding Instructor

Superprof

  • Delivered 30+ remote and in-person lessons in Python, web development, and mobile app development
  • Designed custom lesson plans and coding projects matched to each student's pace and goals
  • Mentored 15+ students through debugging, problem-solving, and full project builds
  • Earned a 95% positive student feedback rating using interactive coding tools
8b07282Jan 2019 - Jun 2020Remote

Full Stack Engineer

RHOGIC

  • Rebuilt the company website with server-side rendering, boosting page load speed by 30%
  • Designed and prototyped web and mobile UI components in Figma, improving design consistency
  • Wrote, tested, and reviewed code with the lead developer, keeping applications maintainable and scalable
  • Diagnosed and resolved mobile app bugs, achieving a 25% reduction in crash reports
f1dc7b0Sep 2016 - Mar 2019Abuja

Full Stack Engineer

Liano

  • Built and shipped internal web tools end to end — interface, API, and data layer — replacing processes the team had been running by hand
  • Hardened the stack against repeatable weaknesses — weak auth flows, exposed ports, outdated services — cutting recurring security issues by 35%
  • Wrote Python and Bash automation that cut manual scanning time by 25% and made routine checks consistent run to run

The stack

Tools I reach for when there's a workflow to automate or a platform to ship.

[languages]

JavaScriptTypeScriptPythonSQLSolidityHTML/CSS

[full_stack]

ReactNext.jsNode.jsExpress.jsFastAPIReact NativeTailwind CSSWebSockets

[data_cloud]

PostgreSQLMongoDBFirebaseAWSGCPDockerCI/CDGit

[blockchain]

Hardhatethers.jsweb3.jsIPFS

I teach this too

AI Automation Engineering — From manual workflow to production automation — build systems that do the work.

01Week 1

Automation Thinking

Learn to see workflows the way an automation engineer does: spot the manual work worth killing, map it as input → process → output, and decide when a problem needs AI versus plain code.

02Weeks 2-3

Python & JavaScript for Automation

The two languages that run the automation world. Scripting, data wrangling, and the package ecosystems that let you stand on other people's shoulders.

03Weeks 4-5

APIs, Webhooks & Data Plumbing

Automation is mostly connecting systems that were never designed to talk. REST APIs, authentication, webhooks, and scheduled jobs — the plumbing behind every pipeline.

04Weeks 6-7

LLM Fundamentals & Prompt Engineering

How large language models actually work — tokens, context windows, and why prompts are functions. Write prompts that return structured, machine-readable output every time.

05Weeks 8-9

Building with the Claude API

Your first real AI integration. The Messages API, tool use, and streaming — wrapped in an endpoint that classifies real input for a real workflow.

06Weeks 10-11

Automation Pipelines

Single calls become systems: multi-stage pipelines that ingest, parse, score, and verify. Queues, retries, and idempotency — the difference between a demo and production.

07Weeks 12-13

Human-in-the-Loop & Quality

Full automation is a myth — great systems put humans exactly where they belong. Review checkpoints, consensus scoring, evals, and guardrails that keep automation trustworthy.

08Weeks 14-16

Capstone — Ship an Automation

Scope, build, and deploy a real automation product end to end: a working pipeline with an AI core, a human checkpoint, and a live URL you can put in your portfolio.

$ Sixteen weeks, ending in a deployed automation with an AI core and a human checkpoint. Enrolled students pick up where they left off in the student portal.

Academic background

2024 - 2025

Post-Graduate Certificate in Blockchain & Back-End Development

York University

Toronto, Canada

View Certificate
2020 - 2022

Master's in Computer Science (AI)

University of Jos

Jos

Verified Credential
2015 - 2019

Bachelor's in Computer Science

ABU Zaria

Kaduna

Verified Credential
Building Agentic AI Systems for Developers
Microsoft Azure AI Essentials Professional Certificate
Building AI Products: Security Professional Certificate
Integrating Generative AI into Business Strategy
Google Project Management Certificate
Atlassian Agile Project Management Professional Certificate
Career Essentials in GitHub Professional Certificate

What people say

Anslem demonstrated exceptional technical expertise and professionalism throughout our project. His ability to understand complex requirements and deliver robust solutions exceeded our expectations. A true asset to any team.

A

Ali Mohammadi

Chair Massage ToGo

Working alongside Anslem was an incredible experience. His deep knowledge of full-stack engineering and AI-powered product development, combined with his collaborative spirit, made him an invaluable team member. He consistently delivered high-quality code and was always willing to help others.

M

Mathew

Full Stack Engineer at Rework

$ ./connect --with anslem

Got a workflow that shouldn't need a human?

I'm open to new opportunities — AI automation, full stack platforms, or the messy problems in between. Tell me what you're building.

anslebieg7@gmail.com