Responsible AI that helps
your clients navigate,
decide, and act.
I help impact-driven organizations find where AI can advance their
mission — and build the AI advisors that make it real: intelligent
tools that guide the people you serve through complex decisions,
services, and choices.
I work with social enterprises - nonprofits and for-profit
organizations whose mission creates positive change — in climate,
community, health, and beyond. If the work matters, let's talk.
🌱
Nonprofits
Mission-driven organizations serving clients who need clear,
personalized guidance through complex services and decisions.
⚖️
B Corps & Impact Businesses
For-profit companies with a genuine commitment to social or
environmental impact as a core part of their model.
🌍
Climate & Environment
Startups and organizations helping individuals and households
reduce their environmental footprint.
🏘️
Community & Health
Organizations working at the intersection of public health, social
equity, and community resilience.
What I Do
I help impact-driven organizations in two ways: finding where AI can
advance their mission, and building the AI advisors that make it real.
Some organizations come with a concrete use case already in mind and
are ready to build. Others are earlier in their thinking and benefit
from an assessment and strategy session first. Both are valid starting
points.
Assessment and strategy
Many organizations have a general sense that AI could help them — but
not yet a clear picture of where, how, or whether the risks are
manageable. I help them get there: reviewing their mission and goals,
analyzing their current situation and capacity, and identifying the
highest-leverage opportunities for AI. A typical result is a concrete
use case and a clear recommendation for next steps.
A direct-service nonprofit came to an initial strategy session
with a general sense that AI could help them — but without a
clear picture of where, how, or whether the risks were
manageable. Like many organizations their size, their staff had
encountered generative AI through consumer tools like ChatGPT,
and had real questions: Would client data be safe? Would AI give
wrong or harmful answers? How much would it cost to maintain?
Was this even realistic for an organization without a dedicated
IT team?
We started by reviewing their mission and the specific goals
they were trying to advance — not AI goals, but program goals:
reaching more clients, reducing the time staff spent on routine
guidance, making their services more accessible to people who
didn't know where to start. Then we looked honestly at where
they stood today relative to those goals, and conducted a SWOT
analysis focused on their capacity to benefit from AI. That
surfaced something they hadn't fully seen: a substantial body of
program knowledge that existed in documents and staff expertise
but hadn't been digitized in a form an AI advisor could use. It
also surfaced an opportunity they had been thinking too small
about — their intake process, which staff found repetitive and
clients found intimidating, was a natural candidate for an
AI-assisted first touchpoint.
By the end of the session, we had identified a concrete use
case: an AI advisor that would guide new clients through the
intake process, ask the right questions, explain available
services in plain language, and hand off to a staff member with
a structured summary — reducing burden on both sides. We
discussed the ethical questions directly: how the advisor would
identify itself as an AI, how it would handle sensitive
information, when it would defer to a human. Having a concrete
use case made those conversations productive rather than
abstract. (For a fuller account of how I approach these
questions in every engagement, see
Responsible AI in Practice.)
The session ended with a recommendation to move into Explore —
and enough clarity to know exactly what the prototype should do.
Building AI advisors
Unlike a chatbot that merely answers questions, a well-designed AI
advisor understands context, adapts to individual circumstances, and
guides people through complexity the way a knowledgeable colleague
would. It reflects your organization's expertise — and extends your
reach.
I bring that depth directly — from AI research through engineering
leadership — so the expertise you need comes with the engagement, not
as a separate problem to solve.
Sarah owns an older home and has been meaning to do something
about her carbon footprint for a while — she just hasn't known
where to start. She opens the advisor on her laptop one evening
and answers a few quick questions: she owns her home, it was
built in the 1970s, and she's looking for a few quick wins
before tackling anything major. When asked what matters more to
her — saving money or reducing her environmental impact — she
says the environment comes first, though she wouldn't mind
saving money too.
The advisor starts her with the easy stuff: switching to LED
bulbs, adjusting her thermostat schedule, sealing drafts around
windows. Then it asks a question that surprises her a little:
"Do you have a gas water heater?" She does — it came with the
house. The advisor explains that for a home like hers, a gas
water heater is typically one of the largest sources of
emissions, and recommends replacing it with a heat pump water
heater. It links to available rebates in her area and estimates
she'd save several hundred dollars a year on energy bills.
She hadn't thought about her water heater. She came in expecting
to change her light bulbs. She left with a prioritized action
plan and several clear next steps she was actually ready to
take.
Full-scope engagement
Every aspect of a transformation toward a responsible AI-enabled
service is in scope: strategy, data preparation, system design,
implementation, and go-to-market — whatever your organization needs.
Responsible AI in Practice
Building AI for the populations nonprofits and impact organizations
serve carries real responsibility. The people interacting with an AI
advisor may be navigating health decisions, financial choices, housing
options, or access to services they urgently need. Getting it wrong
isn't just a product failure — it can cause harm.
Here is how I approach that responsibility concretely.
Data and privacy
No sensitive data stored in the model — and gaps in your data
policies are flagged before we build.
Sensitive information shared in a session is stripped of
direct identifiers before it reaches the AI. No sensitive data
is stored inside the AI module itself. All data handling
follows the organization's own policies — and if those
policies have gaps, I'll flag them and help connect you with
the right expertise before we build. When using AI platforms
via API — my default — prompts and completions are not used
for model training and are not retained beyond a short
processing window.
Bias and transparency
Tested against real use cases before launch, with full
transparency to users about what they're interacting with.
AI can make mistakes, and bias is possible in any AI system.
Every deployment is tested across typical user profiles and
use cases before launch — a process the organization
participates in, since they know their clients best. Users are
always told they are interacting with an AI, and are invited
to flag responses that feel wrong. Where the advisor draws on
the organization's own knowledge base, that is made visible:
"this comes from our resources" is a different kind of
recommendation than "this is my general understanding," and
users should know the difference.
Keeping humans in the loop
The advisor informs decisions — it doesn't make them. High-stakes
topics always route to an expert.
AI advisors are informational tools, not decision-makers.
Recommendations are always framed accordingly, and users are
encouraged to verify before acting. For high-stakes decisions
— health, legal, financial — the advisor explicitly encourages
engagement with an expert, whether at the organization or
elsewhere. In many deployments, the advisor is designed to
inform an expert's decision, not replace it.
Guardrails and scope
Explicit limits built with the organization — and a clear path to
a human whenever the advisor reaches its edge.
Every advisor is built with explicit constraints developed in
collaboration with the organization: what topics are in scope,
what the advisor should do when it reaches the boundary of its
knowledge, and how it handles situations it isn't equipped
for. A well-designed advisor knows its own limits — and always
has a clear exit ramp to a human, a resource, or a next step.
Generic constraints (such as avoiding recommendations that
could harm the user or third parties) are part of every build,
reviewed with the organization before launch.
What's in scope by default — and what isn't
Mobile-first by default. Multilingual support and WCAG compliance
are available when your clients need them.
Multilingual support and formal accessibility compliance
(WCAG) are available but not part of the default engagement
scope. If your clients need either, we scope it explicitly.
All interfaces are designed to be responsive and work well on
mobile.
How to Work with Me
Explore. Design. Build.
I offer a structured process designed to minimize risk, build shared
understanding, and deliver something real at every stage.
01
Phase One
Explore
⏱ Two weeks📅 Up to 3 meetings
We learn about your organization, your clients, and the problem
you want to solve. By the end, you'll have a clear picture of
what's possible — and a working prototype to react to. For most
organizations, the real question isn't how to implement AI — it's
whether it's worth pursuing at all. Two weeks and a working
prototype answer that question concretely, before any meaningful
investment is required.
Discovery interviews with your team
Review of your services, data, and client profile
Initial assessment of opportunities and constraints
A working mockup or proof of concept
Free — no commitment
02
Phase Two
Design
⏱ One month📅 Up to 2 meetings/week🕐 Up to 20 hours
This is where we figure out — together — what your AI advisor
should do, and just as importantly, what it shouldn't. We work
through your real constraints, test our assumptions, and build a
plan grounded in your organization's actual situation — one you
can execute with confidence.
AI advisor architecture and data requirements
User experience design and conversation flows
Integration and platform strategy
Implementation roadmap with milestones
In-kind for nonprofits · Fixed fee for for-profits
03
Phase Three
Build
⏱ Open-ended📅 Flexible engagement
I help you bring the plan to life — as much or as little as you
need. From technical implementation to team training to launch
support, we build together.
AI system development and integration
Data transformation and pipeline work
Testing, iteration, and quality assurance
Launch support and go-to-market guidance
Hourly rate
Work
A few examples.
Climate · Startup
Conversational Climate Action
A startup with a database of household emissions-reduction actions
needed a way for users to explore options that felt personal, not
like a spreadsheet.
Within days of our first conversation, I had built a working
prototype of an AI advisor that lets users explore actions —
replacing gas appliances, changing energy sources, adjusting
daily habits — in a conversational, personalized way. The
advisor understood individual circumstances and helped users
move from awareness to commitment, surfacing next steps and
implementation resources as they went.
Phases involved: Explore, Build
Circular Economy · Startup
AI in the Loop for Resellers
A startup helping resellers bring used items to market faster
wanted to apply AI to improve product descriptions and selling
velocity.
Through careful analysis of a seller's journey on reseller
marketplaces, we designed processes and user interfaces and
applied machine learning and image classification algorithms
that significantly improved the quality of product placement.
This measurably increased seller satisfaction and sales volume
for small to medium-sized resellers.
Phases involved: Design, Build
Community · Nonprofit
AI Strategy for a Nonprofit
A nonprofit wanted to understand where AI could create the most
value — without wasting time or resources on the wrong things.
A direct-service nonprofit had a general sense that AI could
help them serve more clients and reduce staff burden — but no
clear picture of where to start or whether the risks were
manageable. A strategy session surfaced what wasn't obvious at
first: a wealth of program knowledge that hadn't been
digitized, and an intake process that was exhausting staff and
intimidating clients. They left with a concrete use case,
answers to their most pressing questions about data and
ethics, and a clear recommendation for next steps.
Phases involved: Assess, Explore
Philantropy · Startup
Collaborative Portfolio Selection
A startup supporting people in their philatropic giving wanted to
provide a collaborative chatbot experience that helped put
together a portfolio of nonprofits to match their priorities.
We designed a collaborative chatbot that asked the right
questions provided thoughtful suggestions, and interactively
built a targeted portfolio matching the user's goals. Through
careful prompt engineering, constrained guidance, and visual
and natural language interaction, it became a powerful
companion for navigating the 1.8M nonprofits in the startup's
database.
Phases involved: Explore, Design, Build
About
I believe the most powerful AI systems are the ones that put
people — real people, with complex lives and real decisions — at
the center.
My approach begins with deep listening: understanding your
organization's mission, your clients' needs, and the specific
friction points that AI can genuinely address. A prototype early,
a clear strategy before we build, and honest guidance throughout.
I work with a small number of organizations at a time, which means
you get my full and personal attention. For implementation, I work
with your existing team where possible — and can bring in trusted
partners in design and engineering where needed.
Started as a scientist with a PhD in Artificial Intelligence,
developing real-time intelligent systems
Held senior management positions, including VP and Director,
Intelligent Systems Lab, at Xerox PARC
Gained deep engineering experience as startup CTO across
technology and impact sectors
Focusing exclusively on social enterprises with a mission worth
scaling
Contact
Let's explore what's possible.
The first step — the Explore phase — is always free. Tell me a little
about your organization and what you're hoping to solve, and I'll be
in touch within a few days.
