Senior UX Leader — Portfolio Athlone, IE · Bangalore, IN · Remote
Available for new roles

Enterprise software, designed for the AI-native era.

Twenty years leading UX for Ericsson, HP and Ellucian — now designing — and shipping — agentic software: systems where people direct AI with confidence and control.

Portrait of Mohan Venkatesh
Mohan Venkatesh — Bangalore → Athlone
EricssonHPEllucianDiagnox
Scroll
01 — Flagship

Prompt to production — generative UI at Ericsson scale.

The EDS MCP AI Agent — an agentic system I led the UX for at Ericsson: developers describe the interface in plain language; the agent builds it to design-system spec, with a human in command.

Developers spent days hand-building dashboards against the Ericsson Design System — and still shipped non-compliant UI that design teams had to rework.

The design problem wasn't code generation — it was trust. An agent that silently writes UI is a liability; an agent developers can direct, inspect and correct is a force multiplier. Three decisions shaped the experience:

  1. The human stays in command

    Every generation is a proposal. Developers review, refine or reject before anything lands — no silent automation, ever.

  2. Ground the agent in the design system

    The agent reads live component specs from Figma and validates its own output against EDS rules — compliant by construction, not by review.

  3. Design for iteration, not one-shot magic

    Conversational refinement — “make the table filterable” — replaces prompt-perfectionism, matching how developers actually work.

Outcome

A single prompt now produces a complete, EDS-compliant dashboard — 100% design-system components, zero custom CSS — with the developer approving every step. Hours of hand-built UI work replaced by a directed, inspectable generation loop.

Generated in one pass — illustrative recreation; product UI is under NDA
02 — Independent Builds · 2025–26

Seven AI products, shipped solo — directed, not typed.

After twenty years designing enterprise software, I spent 2025–26 proving a thesis: a UX leader who understands the full SDLC can direct agentic AI to build production software — if, and only if, the human stays in command. Seven products are live. Every claim below is verifiable in a deployed system.

  1. Adversarial review gates — MAFIP

    Every feature slice passes an independent AI review scored /100; below 95, or any critical finding, it does not ship. A flagship feature gated at 91 was remediated to 97 before release. Two sub-standard slices were held back entirely.

  2. Evaluation infrastructure AI code must pass

    A 340-test backend suite, golden-reference regression charts, Playwright E2E with axe accessibility checks and 21 visual baselines, CI quality gates. “It compiles” is not a bar.

  3. Design systems under AI generation

    Locked type, color, and elevation tokens with a written taste record — so agent-generated UI cannot drift into generic AI slop.

  4. Human-in-the-loop, across the whole stack

    Conversational AI for thinking, agent skills for automation, MCP for integration, AI IDEs for building — prompt, context and agentic-loop engineering to direct each surface, and verification before anything ships.

The takeaway

The takeaway isn't that AI writes code. It's that judgment scales — when you build the gates that force AI to meet a senior practitioner's bar.

03 — Selected Work

Health, time management and money — three hard problems, one discipline.

Underpinning all of it: formal usability programmes with global telecoms at Ericsson — read one synthesised research story. Full decks under NDA, available on request.

04 — Experience

Twenty years, one throughline: enterprise UX.

Independent AI Product Builder · UX ConsultantSeven production AI products · uvwx.me
Senior UX Designer · Agentic UX LeadEricsson · Athlone, Ireland
UX ConsultantDiagnox · Bangalore
UX Manager — team of 21Ellucian · Bangalore
Lead UX AnalystEllucian · Bangalore
Usability LeadHP · Bangalore

Along the way: introduced atomic design to an entire product suite at Ellucian, and built one of the earliest production agentic UX practices at Ericsson.

20+years designing enterprise software
21designers & researchers led
7AI products shipped to production, 2025–26
100%task completion across 15 required attempts in customer validation
05 — Words

From the people I've worked with.

“He had the ability to focus on the right solution — not the easiest… he introduced atomic design methods to an entire product suite.”

Sania Khan — SVP, Product & Strategic Innovation · Ellucian

“Mohan combines strategic UX thinking with hands-on technical skill, and he's someone I've learned a great deal from.”

Brian Lawless — UI Framework PO · Ericsson

“The work he has done to standardize the common components pattern library has been tremendous.”

Raja Saravanan — Sr Director of Engineering · Ellucian

12 full recommendations on LinkedIn →
06 — Contact

Open to Senior UX Lead & AI-UX roles — permanent, contract or advisory.

Or send a note directly
Healthcare · UX Consultant · 2016 — 2019Diagnox Digital Health

Problem. Diagnox needed one product experience that served patients, clinicians and administrators — and it had to feel trustworthy, usable and clinically credible to all three.

My role. End-to-end UX as independent consultant — stakeholder research, user flows and information architecture, analysis and reporting modules, branding and style guide.

Key decisions. Trust-first patient journeys; analysis and reporting modules designed for clinical insight, not just data display; a single visual system spanning iOS, Android and web.

Outcome. Bridged clinical credibility with consumer-grade usability — the design foundation for Diagnox's platform across its stakeholder network.

Full case study — 8 screens →
Higher Ed · Ellucian · UX Manager · 2015 — 2016Banner Time Entry

Problem. A single time-and-leave flow served employees, supervisors and super-admins equally badly — three personas with very different jobs forced through one screen.

My role. Led the design team; owned persona strategy and the interaction model.

Key decisions. Persona-specific views instead of one-size-fits-all; a calendar view for inherently temporal tasks; team availability surfaced so supervisors could decide, not just approve.

Outcome. Responsive web application shipped across the Banner higher-education suite.

Full case study — 9 screens →
Higher Ed · Ellucian · UX Manager · 2015 — 2016My Finance Query

Problem. Budget owners depended on finance staff for every budget, encumbrance or payroll question — a human bottleneck on routine answers.

My role. Led design of a self-service query experience.

Key decisions. A create-view-share model that turns one person's query into a reusable team asset; progressive complexity from simple lookups to composed queries.

Outcome. Self-service financial answers without the finance-team bottleneck — shipped in the Banner suite.

Full case study — 10 screens →
Usability programmes · Ericsson · Global telecoms

Research practice — one story

Method. Moderated usability studies and heuristic evaluations across programmes with Tier-1 operators — Telstra, Optus, Swisscom, T-Mobile — on Ericsson Service Orchestration & Assurance. Full decks under NDA, available on request.

What we found. A heuristic evaluation across 200 guidelines and 9 usability categories with 5 UX experts established a 72% baseline and surfaced 28 UX issues — 5 critical, 10 major, 13 minor. A separate evaluative study quantified a 35% task-failure rate across 20 pre-release task attempts.

What changed. Findings were synthesised into roadmap-ready GA fixes: navigation, data-rich table usability, role-based status visibility, advanced filters, exception handling and empty-state messaging.

Result. All 28 issues were closed before GA. Customer validation with Telstra and Optus users achieved 100% completion across 15 required task attempts, and 12 improvements shipped in the GA release.