Toratech

AI training that uses your own work

Four packages, 12 weeks each. Every weekly lab is built on the learner's real job, split by role, so what they practise on is what they go back to on Monday.

1

Prompt Engineering

Anyone whose job is mostly writing, analysing or deciding

Length
12 weeks
Modules
15
Per week
8 hrs
Entry
Computer literacy

What you gain

Works through all 28 prompting techniques on their own job, and becomes the person in the department whose AI work is trusted.

What your organisation gains

Two to four hours a week back per knowledge worker, with output quality measured rather than assumed.

What you build

  • A personal AI assistant set up for your own job and writing voice
  • An email drafting and reply set for the messages you send most
  • A meeting-notes-to-actions workflow
  • A question-answering setup over your own documents, with citations
  • A reusable prompt library your whole team can work from

Three courses

  1. 1.1
    How language models behave
    Weeks 1-3
    • What the model predicts, and why that explains its behaviour
    • Context engineering: what belongs in the window, what to leave out, and why more context is often worse
    • Reasoning effort instead of asking for step-by-step working: what the models now do for you
    • Sampling controls, determinism, and getting the same answer twice
    • Working across English, Amharic, Afaan Oromoo and Tigrinya
  2. 1.2
    Advanced technique
    Weeks 4-7
    • All 28 prompting techniques, worked through against your own job rather than toy examples
    • Which of those techniques reasoning models have made redundant, and what replaced them
    • Grounding in your own documents, with citation and a refusal path
    • Structured output: JSON schema enforcement and constrained decoding
    • Tool and function calling, and the Model Context Protocol
    • Prompt injection, exfiltration, and keeping personal data out of logs
    • Building a personal assistant tuned to your own role and voice
  3. 1.3
    Evaluation and production practice
    Weeks 8-11
    • Golden sets, rubric grading, and LLM-as-judge calibrated against human labels
    • Regression suites that block a change on a measured quality drop
    • Prompt versioning, review, and A/B testing against live traffic
    • Token accounting, caching, and right-sizing the model to the task
    • Knowing when prompting is the wrong tool

Capstone, week 12

One real task from your own work, with a 30-example evaluation set and measured before-and-after accuracy.

12 weeks · 15 modules

Priced per learner or as a closed cohort for your team

Request pricing

Certificate awarded on completion

2

AI for Startups

Founders, product leads, early hires

Length
12 weeks
Modules
13
Per week
9 hrs
Entry
Package 1 or placement quiz

What you gain

A founder who can direct AI work and judge it, instead of outsourcing the judgement to a contractor.

What your organisation gains

A shipped product with real users and a cost per user you can defend to an investor.

What you build

  • A customer support chatbot for your website
  • A Telegram or WhatsApp bot your users can actually reach
  • A lead qualification and routing agent
  • An onboarding assistant that walks a new user through setup
  • A career guidance agent, built as a worked example of a sellable product

Three courses

  1. 2.1
    AI product strategy
    Weeks 1-3
    • Where AI creates value, and the use cases that demo well then die
    • Unit economics: cost per user, margin, and what happens at 10x
    • Defensibility when the model is a commodity: data, distribution, workflow
    • What actually sells to Ethiopian SMEs, NGOs, banks and diaspora buyers
  2. 2.2
    Building the product
    Weeks 4-8
    • Retrieval architecture: chunking, hybrid search, reranking, and measuring recall apart from generation
    • Agent architectures: the tool-calling loop, planner and executor, multi-agent handoff
    • Model Context Protocol: exposing your own systems as tools an agent can call
    • Streaming, latency budgets, and interfaces that stay usable while the model thinks
    • Hosted against self-hosted: cost, control, latency, data residency
    • Tracing and evaluation in place before your first paying customer
  3. 2.3
    Getting paid
    Weeks 9-11
    • Pricing: per seat, per use, per outcome
    • Designing a pilot that converts, and the security questionnaire you will be sent
    • The AI pitch, and demoing honestly

Capstone, week 12

A shipped product with at least five real users, a cost model and a pilot deck.

12 weeks · 13 modules

Priced per learner or as a closed cohort for your team

Request pricing

Certificate awarded on completion

3

AI for Enterprise

Department heads, IT and digital leads, risk and compliance

Length
12 weeks
Modules
15
Per week
8 hrs
Entry
Package 1 recommended

What you gain

A department head or chief engineer who can sponsor, govern and defend AI adoption.

What your organisation gains

A costed, governed rollout instead of staff quietly pasting company data into personal accounts.

What you build

  • An internal knowledge assistant over your policies and procedures
  • An HR policy and benefits assistant for staff
  • An IT helpdesk triage agent
  • A document review assistant with human approval gates
  • A management reporting assistant drawing on your own numbers

Three courses

  1. 3.1
    Strategy and governance
    Weeks 1-4
    • Opportunity mapping scored on value, risk and effort
    • Acceptable use, approval gates, and the shadow AI already in the building
    • Model risk management: documentation, sign-off, audit trail
    • Data readiness: inventory, classification, residency, retention
    • The regulatory frame: data protection duties, NBE directives, the EU AI Act as a reference
  2. 3.2
    Architecture, security and operations
    Weeks 5-8
    • Three reference architectures: document assistant, processing pipeline, agentic workflow with approval gates
    • Model routing and fallback chains: cheap model first, escalate on difficulty
    • Private deployment: open-weight models, quantisation, GPU sizing, air-gapped operation
    • Observability: distributed tracing across retrieval, model and tool calls
    • Red-teaming: the OWASP LLM top ten, indirect injection, exfiltration through tool calls
    • Drift detection, and cost governance charged back per department
  3. 3.3
    Adoption and measurement
    Weeks 9-11
    • Rolling out: pilot, department, organisation
    • Role-based enablement, so each function learns only what it needs
    • Measuring return honestly: baseline before, not after
    • Job redesign, and what meaningful human oversight actually requires

Capstone, week 12

A board-ready adoption plan: opportunity map, one costed pilot with threat model, governance policy, 12-month roadmap.

12 weeks · 15 modules

Priced per learner or as a closed cohort for your team

Request pricing

Certificate awarded on completion

4

AI Automation

Operations, finance, HR, admin, support and engineering staff

Length
12 weeks
Modules
14
Per week
8 hrs
Entry
Computer literacy

What you gain

Staff who can automate their own function. The highest-leverage skill a practitioner can add this decade.

What your organisation gains

Named recurring processes automated, with measured hours saved and error rates before and after.

What you build

  • An email triage and reply agent for a real inbox
  • A calendar and scheduling assistant
  • An invoice and receipt extraction pipeline with a review queue
  • A recurring report that writes itself
  • A personal assistant for your own workload, in your own lane

Three courses

  1. 4.1
    Finding and building the automation
    Weeks 1-4
    • Mapping a process as steps, inputs, outputs and exceptions
    • The exception-rate test, and the work that must not be automated
    • Document, communication and data work, end to end
    • OCR on scanned and Amharic documents, with confidence thresholds
  2. 4.2
    Agents and orchestration
    Weeks 5-9
    • Agent design: tools, scopes, step budgets and a cost ceiling per run
    • Model Context Protocol: exposing your own systems as tools an agent can call
    • Durable workflows: state, retries, and resuming after a failure
    • Running the same job twice without paying an invoice twice
    • Human in the loop: confidence thresholds, review queues, approval gates
    • Tracing a run end to end, and noticing when it has silently stopped
  3. 4.3
    Sector automation
    Weeks 10-11
    • Build against your own industry's documents and regulator
    • Automotive and EV, construction, power, banking, public service
    • Logistics and customs, telecom, health, manufacturing, agriculture

Capstone, week 12

One real recurring task from your job automated, with a two-week run log, measured hours saved and a handover document.

12 weeks · 14 modules

Priced per learner or as a closed cohort for your team

Request pricing

Certificate awarded on completion

Taken individually or as all four over 12 months, for one learner or as a closed cohort for your team. Tell us which and we will price it.

Role lanes

One cohort shares the live sessions; the weekly lab splits by job function. Engineering lanes take one extra module on verification, because engineering work gets signed, stamped and built.

Business lanes

  • Finance & accounting

    Invoices, reconciliations, variance commentary, audit queries

  • HR & administration

    Job descriptions, policy drafting, onboarding packs, minutes

  • Sales & marketing

    Proposals, campaign copy in four languages, call summaries

  • Customer support

    Ticket triage, reply drafting, escalation rules, quality review

  • Operations & supply chain

    Purchase orders, supplier correspondence, shift reports

  • Management

    Board and donor reports, performance summaries, decision memos

Engineering lanes

  • Automotive & EV

    Fault trees and diagnostic codes, service manuals and wiring diagrams, repair estimates, warranty packs, technician job aids

  • Electrical & power

    Standards lookup, outage and fault reports, load and protection documentation, commissioning records

  • Civil & structural

    Bills of quantities, specifications, method statements, site diaries, inspection reports

  • Mechanical & manufacturing

    Maintenance logs, root cause analysis, FMEA, work instructions, spares and procurement

  • Chemical & process

    Batch records, standard operating procedures, deviation reports, HAZOP preparation

  • Telecom & network

    Fault report clustering, configuration review, capacity and coverage reporting

  • Software & IT

    Code review support, test generation, documentation, incident write-ups

  • Water, environment & agricultural

    Monitoring data, environmental impact reporting, permits, maintenance scheduling

Verification, for engineering lanes only

Calculations are never delegated. Standards and codes are retrieved and cited by clause, never recalled, because a model will produce a plausible clause number that does not exist. The module also covers units and tolerances, what vision models misread on a schematic, safety-critical boundaries and professional liability, and how to record AI assistance in work that will be signed.

Training a team rather than yourself?

Closed cohorts run on your calendar, in your lanes, against your own documents. Most organisations start with one department and one measured process.

Ask about a cohort