Job search
Instructional Design Portfolio for Tech Roles: Show Learning Decisions and Evaluation
Show how objectives, technical practice, feedback, and assessment fit together, with clear ownership and honest boundaries around learning results.
An instructional design portfolio for tech roles should show what learners need to do, how the learning experience prepares them, and what evidence would demonstrate success. Include a finished exercise with feedback and an explanation of your design decisions. A polished course screen or a clear technical article cannot, by itself, show whether learners can apply the skill.
Choose examples that match the role’s audience and technical depth. Developer education, customer training, and internal software onboarding involve different prerequisites and working conditions. Use LandOffer.ai to explore roles, then read the employer’s actual portfolio instructions. This guide uses original fictional learning cases; it reports no learner study, course delivery, or measured improvement.

Explain the learning problem before the course format
Start with the performance you want to support. “Teach configuration” is too broad to explain a design choice. “Given this configuration contract, preserve an explicit zero while resolving an inherited value” identifies an action you can practice and assess. State what the learner already knows and the conditions under which they should perform it. Specify whether they can use a reference, inspect supplied data, or work without prompts; those conditions affect what an answer can demonstrate.
Check whether training addresses the actual problem. CDC’s needs-analysis guidance, dated February 5, 2025, distinguishes knowledge or skill gaps from problems that may need changes in technology, policy, or procedures. Applied to a technical portfolio, a broken practice environment is not proof that the learner needs a longer explanation. Explain the evidence for a skill gap and identify any assumption still needing investigation. Describe an environmental blocker as a blocker rather than diagnosing a learner from an unsuccessful attempt.
Identify your own work: needs analysis, objective writing, exercise design, technical review, feedback, accessibility decisions, delivery, or evaluation. Name collaborators’ responsibilities. If an engineer reviewed the technical contract but you designed the practice and assessment, explain that division rather than claiming the entire course was your independent work.
Connect the objective, practice, and assessment
Carnegie Mellon’s Eberly Center explains alignment among learning objectives, instructional activities, and assessments. A portfolio can make that connection visible through a small complete lesson. If the objective asks learners to diagnose a problem, a quiz that only asks them to recognize a definition is insufficient evidence of that performance.
Technical writing and instructional design can overlap. A reference page may explain an API correctly; a learning design also decides what the learner should attempt, how much support to provide, what errors reveal, and what feedback follows. Include enough of the reference material for the exercise to be self-contained, but make the learning decisions inspectable.
Choose a manageable technical scope. A short text-based exercise can demonstrate judgment without a hosted course, account, or production sandbox. Explain why its format fits the objective and constraints. A reviewer should be able to see why the learner must make a decision or produce an answer, rather than only watch an explanation. Do not claim that a particular authoring tool or interaction is required for every technology education job.
Worked case: configuration inheritance without losing zero
This case is entirely fictional. Designer Tamsin Hale creates a short lesson for developers who can read simple objects and conditionals. The invented configuration system is not a real product, and its rules must not be applied to another system without checking that system’s contract.
Learning objective: Given the rules and four inputs below, select the effective timeout or identify a rejected input, and explain why an explicit zero is preserved.
Design constraint: Learners have a text-only practice environment. The lesson cannot depend on a runnable service or screenshots. Tamsin owns the exercise, feedback, and assessment design; a technical reviewer would need to approve the contract before any real delivery. No such delivery is claimed here.
Finished learner instruction: “Resolve timeout_ms in this order: request, project, then default. The default is 600. An absent value or null means inherit from the next level. An integer zero is a supplied value and means no timeout in this fictional system. A non-integer value, including an empty string, is rejected; do not silently inherit from it. First check whether the request value is supplied and acceptable, then consider the project value if inheritance is required.”
Practice prompt: “For each case, give the result and one sentence explaining the decisive rule. Do not rewrite the inputs.”
- Case A: Project value is
0; request value is absent. - Case B: Project value is absent; request value is
900. - Case C: Project value is
250; request value isnull. - Case D: Project value is
250; request value is an empty string.
Correct answer and rationale: A resolves to 0: the request inherits, and the project explicitly supplies zero. B resolves to 900: a supplied request integer takes precedence. C resolves to 250: request null inherits the supplied project value. D is rejected: an empty string violates the stated input rule and is not an inheritance marker.
Plausible wrong answer: “A is 600 because zero means nothing was configured.” This reveals confusion between a supplied zero and an absent value.
Feedback shown for that answer: “The contract gives zero a meaning: no timeout. The request is absent, so inspect the project value. It is the integer 0, which is supplied and acceptable. Keep 0; do not move to the default. Try explaining which two values mean inherit.”
Further feedback: For D, “250” would skip input validation. The response is: “An empty string is rejected under this exercise’s rules. Inheritance applies to absent or null values, not every value that seems empty.” This feedback targets the mistaken rule rather than merely displaying a score.
Independent assessment item: “Project value is null; request value is 0. What happens?” Correct response: 0, because the supplied request integer takes precedence. An answer of 600 is consistent with treating zero as missing, but the learner’s explanation is needed to distinguish that mistake from a precedence error; 250 would import a value from the earlier practice rather than reading the new input.
Evaluation plan: Ask learners for an explanation as well as the result. Record errors separately for precedence, inheritance markers, and validation. Use a different but equivalent input set before and after instruction if the purpose is to investigate change in understanding. Review any item whose wording, rather than the target concept, appears to cause confusion. These are proposed activities, not collected results.
Portfolio interpretation: Tamsin can show a finished lesson, targeted feedback, and a distinct assessment item. She cannot claim that developers stopped shipping configuration bugs, completed a course, or improved their scores. The next step would be technical review and an appropriately planned learner pilot, with any findings reported separately.

Worked case: count records without hiding missing measurements
The second fictional case concerns SQL interpretation for support analysts who already know SELECT and GROUP BY. Designer Merrick Vale wants learners to distinguish the number of records from the number with a recorded duration. This is synthetic training data, not customer telemetry or a production investigation.
Learning objective: For the supplied dataset, write a grouped query that retains every record and separately counts recorded durations, then explain how missing values differ from zero.
SQLite’s official aggregate-function reference documents the count distinction used below. The example concerns these basic counts, not every SQL dialect or production data rule.
Finished learner reference: “In this exercise, COUNT(*) counts rows. COUNT(duration_ms) counts rows whose duration is not NULL. A recorded zero is not NULL. Keep all rows when comparing total records with recorded measurements; filtering missing durations before grouping would remove part of the evidence.”
Dataset: The table is called sample_runs and contains exactly these six rows:
alpha, duration120alpha, duration0alpha, durationNULLbeta, durationNULLbeta, durationNULLbeta, duration80
Practice prompt: “Return each team, its total record count, and its recorded-duration count. Then state how many records have a missing duration for each team. Do not replace missing durations with zero.”
Complete correct query:
SELECT team,
COUNT(*) AS total_records,
COUNT(duration_ms) AS recorded_durations
FROM sample_runs
GROUP BY team
ORDER BY team;
Correct answer and rationale: alpha has three total records and two recorded durations, leaving one missing duration. The zero counts as recorded. beta has three total records and one recorded duration, leaving two missing durations. These counts describe this supplied dataset; they do not diagnose why a measurement is missing.
Plausible wrong solution: Add WHERE duration_ms IS NOT NULL before grouping, then report the resulting totals as all records. That produces two retained rows for alpha and one for beta, hiding the missing-duration records.
Feedback: “Your filter removes the rows needed to compare completeness. Keep the full dataset and use the two counts for different questions. For alpha, the rows are 120, 0, and NULL: three records, two recorded values. Revise the query without the filter and explain why zero still counts.” A second wrong answer, “alpha has only one recorded duration,” receives feedback to inspect the explicit zero rather than treating it as missing.
Independent assessment: “A new team has durations NULL, 0, 0, and 50. Give total records, recorded durations, and missing durations.” Correct response: four total, three recorded, one missing. The repeated zeros make it harder to succeed by memorizing the earlier output.
Design rationale and constraints: Merrick uses a tiny visible dataset so learners can inspect every row. He rejects a multiple-choice definition-only item because the objective includes writing and explaining a query. A technical subject-matter reviewer should check the target database context before this becomes a real course. The exercise does not cover joins, duration validity, or causal analysis of missing data.
Evaluation plan and ownership: Merrick would use three separate criteria: the query retains all records, the two aggregates answer the intended questions, and the explanation counts zero as recorded while excluding NULL. Each criterion is met or not yet met, with feedback tied to the observed error. He would review error patterns to decide whether the reference, practice, or feedback needs revision. A later transfer task could use a different approved dataset without the reference. He owns this learning design; no learner testing, workplace observation, or training result has occurred.
Evaluate learning separately from completion and satisfaction
CDC’s evaluation guidance, reviewed October 28, 2024, separates learning from applying it at work. It also explains that a posttest alone can show end-of-training proficiency without establishing a change from the learner’s starting point. Satisfaction responses do not, by themselves, establish training effectiveness.
In a real portfolio, say what was measured, when, and with which task. A completed module is an activity record. A correct response is evidence about that item under those conditions. Applying the skill later without the same support is a different question. Neither a passing answer nor a completion record establishes competence across unrelated tasks or readiness for a job. Report sample size and limitations when you have authorized results, and explain concurrent changes that could affect an interpretation.
A planned evaluation is still useful evidence of your design thinking. Label it as proposed, identify the decision it would inform, and state what you would change if learners struggled. Avoid filling the gap with invented percentages or a claim that a course made people job-ready.
Package a case so reviewers can inspect your judgment
Lead with audience, objective, your role, project status, and constraints. Include enough finished material for a reviewer to attempt the exercise and compare the feedback. Show one rejected approach and explain why it did not fit. Keep supplementary storyboards or facilitator notes available without making them necessary to understand the central case.
Use permitted materials only. Replace private learner records and proprietary product details with clearly labeled reconstructions or original exercises when appropriate; do not present them as real delivery evidence. Check that any diagrams, code, captions, and answer keys remain readable in the requested submission format.
The public guidance here was accessed October 22, 2026. Carnegie Mellon’s reviewed page does not display a publication date; the CDC dates above identify the guidance, not a new hiring rule. None establishes employer acceptance, course accreditation, or a credential guarantee.
To tailor your selection, browse recent roles on LandOffer.ai and choose an audience and responsibility set to compare with your cases. Choose the case that explains those learning decisions. Before sharing it, check the employer’s requested format and label what was designed, delivered, and evaluated using only evidence you can support.


