hi, I'm Aditya
Founder · Builder · Technical leader
Turning complexity into possibilities.
Founder and technical leader. Engineering is home base; product, marketing, sales and hiring are the other hats, often in the same week. For six years I've been turning hard, human problems into things people actually use.
Things I keep saying
Same curiosity. Bigger problems., Hard problems usually start as a conversation., Ship small. Learn fast. Write it down., Latency is a feeling, not a number., Good teams make each other braver.
(usually building something)
Ideas → Systems → Products → Teams → Impact
From curious ideas to real impact.
The part I enjoy most is the whole arc: a question nobody has answered well yet, a scrappy first version, and the day real people start relying on it.
- 01Idea
- 02Explore
- 03Build
- 04Ship
- 05Impact
Beyond the code
The work around the work.
Starting companies from the first commit teaches you that the code is only part of it. Somewhere between the architecture diagrams there are first hires, roadmaps that have to fit reality, long calls with co-founders, and clinics explaining what really slows them down. I've come to enjoy that part as much as the building.
the partnobody sees
Choosing what not to build
Most of the useful decisions I've made were quiet ones: the feature we skipped, the rewrite we postponed, the problem we picked instead.
Growing people with the product
Early teams are small, so every person's growth shows up in the product. I write things down, review honestly, and try to leave room for others to own things.
Keeping promises to users
A dentist trusting a summary, a rep trusting their score. Reliability is less a metric than a promise someone made on your behalf.
Explaining trade-offs plainly
Whether it's a co-founder, a customer or an investor asking, the best technical answer is usually the one a non-engineer can repeat.
- Two companies from day zero
- First hires
- Roadmaps that meet reality
- Writing things down
- Listening to customers
Along the way
- Humantic AIFounding engineerUnderstanding people through language
- Skit.aiSoftware engineer · Voice AIThree years of conversational voice systems
- DextegoCo-founder & CTOReal-time AI sales coaching, built from zero
- CliniChatCo-founder & CTO, nowAI for dental clinics in Italy
Things I keep relearning
Same curiosity.Bigger problems.
Latency is the product.
A voice that answers a beat too late stops feeling human. I've learned to design around the next few hundred milliseconds.
AI suggests. People decide.
In coaching and in clinics, a person looks at what matters before it reaches anyone else.
Stay close to the whole path.
From the microphone to the interface. Knowing every hop is how surprises get caught early.
Build for the room you're in.
Italian clinics deserved Italian software from the first day, not a translation later.
CliniChat AI
Less admin.More patients.Happier clinics.
AI for Dental Clinics
Co-founder & CTONow
An AI assistant built into the everyday work of a clinic, from the consultation to the summary a patient takes home.
Visit CliniChat- Live consultation transcription
- Clinical AI with doctor review
- Treatment planning
- Dental charting
- AI automation
- Healthcare
- Real impact
CliniChat impact
Care made simpler.
Less time writing up the visit, more time for the person in the chair.
Consultations, transcribed live
The conversation in the chair becomes a transcript while it happens, and survives a dropped connection.
Streaming speech-to-text over WebSockets, reconnect and replay
A summary after every visit
Chief concern, findings and next steps, written up for the clinic and in plain words for the patient.
Treatment plans with a human in charge
AI proposes procedures from a structured catalogue; a multi-stage doctor review decides what reaches the patient.
Grounded reasoning, immutable audit trail
Charting that keeps up
An interactive dental chart in FDI notation, updated live as observations are made during the visit.
SVG canvas, WebSockets, React Query
- ConsultationAudio captured in the clinic
- Live transcriptStreaming STT over WebSockets
- Audio pipelineCelery workers, Redis
- Clinical AIGrounded summaries and plans
- Doctor reviewMulti-stage, fully audited
- PatientClear, approved summary
Dextego
Real-time AI coachingfor real people.
AI Sales Coaching Platform
Co-founder & CTO2025
Real-time AI coaching that helps sellers read the room, adapt, and influence the people who decide.
- Real-time coaching
- AI roleplay
- Buyer insights
- Multi-agent coaches
- Sales
- Influence
- AI agents
- Coaching
- Revenue teams
Dextego deep dive
From roleplay to real revenue.
Practise before the call, get coached during it, and learn from it afterwards.
Practise with an AI buyer
Sellers rehearse the hard call against an AI prospect that pushes back, interrupts and changes its mind.
WebRTC SFU for low-latency audio and video
Guidance during the call
Prompts arrive while the conversation is still happening, not in a review a week later.
Go + FastAPI services over gRPC, Redis
Coaches with separate jobs
Distinct AI personas for roleplay, questions and motivation, each with its own role and memory.
Multi-agent LLM orchestration
Know the buyer first
Signals from several sources fused into one read of who is in the room and what they care about.
Context-synthesis service
- CallBrowser or meeting audio
- WebRTC SFULow-latency media
- OrchestrationGo + FastAPI over gRPC
- ContextSignals fused per buyer
- CoachesMulti-agent personas
- GuidanceLive prompts, then feedback
Deal Drive
You move on.The deal doesn't stall.
Your AI Deal Agent
FounderBuilding
The meeting ends, the agent takes over: a shareable deal case, answers for every stakeholder, and a live read on momentum.
Visit Deal Drive- Deal case, built for you
- Grounded answers
- Buyer stories
- Momentum signals
- Sales
- AI agents
- Buyer experience
Deal Drive deep dive
The deal keeps moving, even without you.
An agent that works for the seller long after the call: building, answering, and watching every stakeholder.
The follow-up, already built
Meeting notes, business case, executive summary and next steps, assembled into one deal case.
Capture → analyse → build
Answers grounded in the deal
Buyers ask, the agent answers from the conversation, the deal case and uploaded material, and says so when it doesn't know.
Retrieval over deal-specific context
A story buyers can share
The deal turned into a short story the champion can circulate to people who weren't in the room.
Momentum you can see
Who opened what, who joined, and the next best action, in real time.
- CaptureConversation and deal context
- AnalyseStakeholders, pains, risks
- BuildDeal case and story
- AnswerGrounded agent replies
- FollowMomentum and next action
Eloquente AI
My firststartup.
Personalised Email Outreach
Founder2024
Hyper-personalised outreach at scale: import a list, research every prospect, and write emails that sound like they were written for one person.
- Bulk personalisation
- Prospect research
- Deliverability
- Campaign analytics
- Outreach
- Personalisation
- Deliverability
Eloquente AI, the first one
Every email, written for one person.
The company is closed now. What it taught me about context, prompts and trust runs through everything I've built since.
Personalised, in bulk
Import leads from a CSV and draft an individual email for every one of them in a single pass.
Batch LLM generation with per-lead context
Research before writing
Each email starts from what can be learned about the person and their company, not from a template.
Online intelligence → prompt context
Lands in the inbox
Connectivity checks and spam mitigation, because a perfect email in the spam folder is a wasted one.
Know what worked
Opens, clicks, meetings booked and engagement per campaign, so the next one is better.
- LeadsCSV import
- ResearchProspect + company context
- DraftingPersonalised by an LLM
- DeliverabilityConnectivity and spam checks
- AnalyticsOpens, clicks, meetings
Under the hood
One engine, two conversations.
At CliniChat it listens in the dental chair. At Dextego it listened to sales calls. At Eloquente it read up on every prospect before writing. Underneath is the same kind of engine I keep rebuilding and refining: understand the conversation with the right context, let models and agents do their part, and hand a person something useful before the moment passes.
Audio from the call or the chair, streamed over WebRTC and WebSockets and transcribed while people are still talking.
The right memory at the right moment: past visits, buyer signals and catalogues, retrieved and grounded before the model sees anything.
Prompts written, versioned and tested like code, with the model told exactly what it may say and in what shape.
Separate agents for separate jobs, a buyer who pushes back and a coach who listens, orchestrated live over gRPC.
LLMs for reasoning today, and a serving layer built so a custom-trained clinical model can drop in as a swap.
Nothing clinical reaches a patient without a doctor. Every AI step leaves an immutable audit trail, and feedback flows back in.
Same engine.New room.
ThinkingIdeas, experiments and perspectives.
Thoughts on AI, product building, engineering and the future of work.
Read My Writing- AI & Society, in progressAI isn't just a feature. It's a new operating model.
- Engineering, in progressWhat I look for in an engineering team.
- Product, in progressWhy most products fail to scale.
Better questions.Real problems.Practical ideas.Measurable impact.
- AI & Society
- Product
- Engineering
- Teams
- Learnings
About
Engineer, product builder, and someone happiest when a difficult problem turns into something simple that people actually use. Usually with a good team around the table.
More About Me- Years building
- 6+
- Startups built from day zero
- 4
- Problems to solve
- ∞
- Working globally
- Open source contributor
- Always learning