A very close friend of mine was using cannabis heavily for nearly three years. What started as recreational slowly became a crutch — and by the time the signs were impossible to ignore, the damage was already visible: lower motivation, emotional flatness, a growing detachment from people he cared about and things that once mattered to him.
He had me around. A support system most people don't have. And even then, quitting was hard. Not just physically — but emotionally. Every bad day felt like a setback. Every slip felt like failure. There was no space to process it without judgement, no tool that understood the nuance of what he was going through.
Watching my friend go through this up close changed how I saw the problem. The turning point wasn't a single moment — it was a slow realisation that there are people out there doing this completely alone. Who don't have a friend to lean on, who can't openly talk about it, who are silently struggling with something society still treats with stigma.
Most existing apps are either clinical and cold, or built for abstinence-only thinking. They don't account for the messy, non-linear reality of trying to change a habit that has deep emotional roots. They don't make you feel seen.
Day One Again is built on one belief: that every single day is a fresh start. That "day one again" isn't a sign of failure — it's proof you're still trying.
Track usage and recovery trends — so users can see their own patterns clearly, without shame. Data as self-awareness, not self-punishment.
Connect with society on their own terms — for those who are ready, a gentle on-ramp back to community and connection, not forced or rushed.
Mark every day one as meaningful — whether it's your first day one or your fiftieth, it counts.
I used ChatGPT to think through structure and content, Claude to refine the experience and writing, and Lovable to bring the interface to life — going from idea to working product without a single line of code written by hand.
What surprised me most was how much product thinking still mattered — knowing what to ask, how to frame the problem, when to push back on what the AI generated. The tools handled the execution. The judgement was still mine.
Sprinklr's Media Monitoring product had effectively become a dead product — losing relevance in a competitive market, unclear on positioning, and being left behind in the GTM motion. The product needed more than a patch. It needed a rethink.
The task was to restructure what the product actually offered, identify where it could genuinely win, and make the UX reflect a product worth selling again.
Started with a deep audit of the existing product, customer feedback, and competitive landscape. Identified the core jobs-to-be-done that customers actually valued versus what the product was trying to be.
Restructured the product offering around a clearer value proposition — simplifying the feature surface, sharpening the positioning, and aligning the roadmap with what the market actually needed.
Revamped the UX to reflect the new structure — reducing cognitive load, improving discoverability, and making the product feel modern enough to sell confidently.
The restructuring had a humungous impact on GTM strategy — giving the sales team a product they could confidently position, a story they could tell, and a reason for customers to reconsider. A product that was effectively written off came back into active pipeline conversations.
Collections is a tricky business by nature — high pressure, highly regulated, deeply human, and notoriously difficult to scale without either burning out agents or compromising compliance. The call centre was running on gut instinct and manual processes. There was no scalable system.
The goal was to productize the call centre operation — to turn what was an art into a repeatable, measurable, scalable system.
Mapped the entire collections calling workflow end-to-end — from lead prioritisation and agent assignment to call scripts, disposition codes, and follow-up logic. Identified where time was being lost, where agents were over-burdened, and where the data was being ignored.
Built a scalable calling framework — productizing prioritisation logic, automating disposition flows, and creating a feedback loop between outcomes and strategy. Made the system smarter over time rather than dependent on individual agent heroics.
The numbers tell the story: 50% improvement in overall calling efficiency, 30% reduction in average handle time, and a direct 25% reduction in agent costs — all while maintaining compliance and improving recovery rates. A system that used to depend on the best agents now worked because the system itself was better designed.
CFM (Customer Feedback Management) is a crowded, commoditized space. Differentiation on features alone isn't enough. The real competitive lever is how easily customers can get up and running — and how little they need to rely on managed services to do it.
The challenge: build a distribution and integration strategy that makes Sprinklr CFM genuinely self-serve at scale.
Leading the digital strategy as PM — designing a self-serve product that lets customers build their own distribution setup with minimal hand-holding. This includes integration architecture, onboarding flows, and the scalable infrastructure that sits underneath.
The goal is to move from a model that requires significant managed services to one where customers can be fully operational independently — reducing cost-to-serve while improving time-to-value.
AI in education is full of hype and short on practical utility. Most tools are built for edtech enthusiasts, not for a teacher with 40 students, a packed curriculum, and no time to learn a new platform. The gap between what AI can do and what actually helps in a classroom is enormous.
Aistra's education charter was about closing that gap — making AI genuinely accessible and tactically useful for the people actually in classrooms.
Led the product charter end-to-end — from understanding how teachers and students actually spend their time, to designing AI capabilities that slotted into existing workflows rather than demanding new ones.
The approach was deliberately two-track: tactical wins that teachers could use immediately (lesson planning, feedback generation, differentiation support), and longer-term infrastructure that built classroom efficiency as a practice — not just a feature.
Worked closely with educators to ensure the tools reflected real classroom dynamics, not idealised ones. Accessibility and trust were as much a design constraint as functionality.
Making AI work in education isn't about the most sophisticated model — it's about the most useful interface. This project reinforced something I believe deeply: the best technology disappears into the workflow. When teachers stopped noticing the AI and just noticed that their job was easier, we knew we were on the right track.