K
AI Platform2025 – 2026HR Technology
This project is delivered under NDA. Client and product identities are withheld; screens are access-gated and shown to prospective clients on request.

AI Assessment & Hiring Intelligence Platform

A multi-tenant enterprise SaaS platform that unifies assessment authoring, AI conversational interviews, resume intelligence, browser-based proctoring, AI answer evaluation and hiring analytics — built AI-first, but engineered to keep working when every AI provider is unavailable.

Product Design Lead & Full-Stack Engineer
2025 – 2026
Enterprise Technology Services
Client:Enterprise Hiring Organisation
Next.js 14ReactTypeScriptFastAPIPythonMySQLGPT-4oVision analysisAzure AD SSO
Problem Statement

Technical assessments, AI voice and video interviews, resume intelligence and proctoring — unified in one multi-tenant platform.

Our Approach

Seven capability pillars behind one application — assessment authoring, AI interviews, resume intelligence, proctoring, evaluation, analytics and enterprise controls — so a candidate's full record lives in one place.

Assessments can be authored manually or generated from a job description, drawing on a reusable question bank spanning multiple-choice, coding, audio, video, scenario and free-text formats.

AI conducts conversational voice interviews and STAR-scored video interviews, while proctoring combines deterministic browser-event integrity scoring with vision-based snapshot analysis.

Every AI call has a deterministic fallback, and speech synthesis falls back through four providers — the platform runs, and demos, with no API keys configured at all.

Challenge

Technical hiring at scale was spread across disconnected tools: one system for assessments, another for interviews, spreadsheets for scoring, and nothing joining them into a defensible decision.

Each additional tool added integrity risk and reduced consistency — two candidates could be assessed on different criteria and neither result could be compared with confidence.

Any platform touching hiring decisions also has to satisfy enterprise reality: multi-tenancy, SSO, audit trails, data-protection obligations and access control.

Solution

A multi-tenant enterprise SaaS platform that unifies assessment authoring, AI conversational interviews, resume intelligence, browser-based proctoring, AI answer evaluation and hiring analytics — built AI-first, but engineered to keep working when every AI provider is unavailable.

Outcome

The pattern applies to any high-volume evaluation process — applications, claims, submissions, audits — where throughput matters but every decision must remain explainable and auditable.

Capability pillars

7, all backed by real modules

Question formats

MCQ, coding, audio, video, scenario, text

Tenancy

Multi-tenant with plan tiers

AI resilience

Deterministic fallback on every call

The constraint that shaped it

It has to work when the AI doesn't

A hiring platform that stops working during an API outage stops a business function. Every generative call sits behind a deterministic fallback and TTS chains through four providers — so the system degrades in quality rather than availability. It runs end to end with no keys configured, which also makes it demonstrable anywhere, offline.

· Decisions that mattered

Deterministic integrity, AI-assisted — not the reverse

Integrity scoring is deterministic and explainable, with vision analysis as an augmentation. An unexplainable integrity flag against a candidate is indefensible, so the explainable signal leads.

Cross-database portability by design

UUID-string primary keys let the same codebase run MySQL in production and SQLite in development, so local development and demos need no infrastructure.

Enterprise controls built in, not bolted on

SSO, RBAC, audit and data-protection requirements were designed in from the start — retrofitting them into a multi-tenant product is where these platforms usually stall.

· What I owned
  • Led product design across all seven capability areas
  • Built front-end and full-stack implementation hands-on
  • Designed how AI assessment surfaces to recruiters so results stay interrogable
  • Defined the fallback architecture that keeps the platform available without AI providers
· Design Process

How We Solved the Problem

01
01

Assessment authoring

Manual or AI-generated authoring from a job description, backed by a reusable multi-format question bank with publish, clone and archive lifecycle.

    02
    02

    AI interview layer

    A conversational voice interviewer plus a video interview module scored against STAR criteria, producing structured, comparable output rather than free-form notes.

      03
      03

      Resume intelligence

      PDF and DOCX parsing with structured extraction, then matching and ranking against the specific role requirements.

        04
        04

        Proctoring & integrity

        Browser event capture feeds a deterministic integrity score, augmented by vision-based snapshot analysis — deterministic first so integrity findings can be explained.

          05
          05

          Evaluation & analytics

          AI grading produces dimension-level scores with calibration, feeding funnels, score distributions and skill breakdowns for the hiring team.

            · Product Screens

            Product Designs

            These are real product screens from a client engagement. Enter the access code to view them.

            NDA Protected

            Concept designs available

            Recruiter dashboard

            Assessment pipeline, candidate flow and outstanding actions in one view.

            NDA Protected

            Concept designs available

            AI assessment authoring

            Generates a full assessment from a job description, editable before publish.

            NDA Protected

            Concept designs available

            Resume intelligence

            Parsed candidate profiles matched and ranked against role requirements.

            NDA Protected

            Concept designs available

            AI video interviews

            STAR-scored video interviews producing structured, comparable output.

            NDA Protected

            Concept designs available

            Integrity & proctoring

            Deterministic integrity scoring from browser events, augmented by vision analysis.

            NDA Protected

            Concept designs available

            Candidate evaluation

            Dimension-level AI scoring a recruiter can interrogate and override.

            NDA Protected

            Concept designs available

            Hiring analytics

            Funnels, score distributions and skill breakdowns across the hiring pipeline.