Ishir Vasavada
Ishir Vasavada
Product leader · Builder · Techstars founder

I turn messy problems into products people rely on.

6+ years across enterprise software, startups and finance. I've owned a $20MM ARR product portfolio, co-founded a Techstars-backed startup, and I still build hands-on, AI included.

See the work ↓ LinkedIn ↗ Resume (PDF)
$20MM
ARR product portfolio I own
$15MM
in deals from a new data-center launch
3,500+
enterprise customers on products I launched
$200K+
raised for the startup I co-founded
Sheet 01 Work

Six products you can try

Things I've built or shipped. Each one runs in the browser.

Agentic rules engine

Based on the agentic rules engine I launched. Describe a billing policy in plain English. An agent turns it into a decision table, tests it, fixes what fails, then waits for approval.

MCP contextevalsguardrails
Who it's for
Billing admins and implementation teams at enterprise subscription companies
Market
Enterprise subscription billing (B2B SaaS)
Problem
Every enterprise bills differently. Custom pricing and decision logic lived in custom code, services projects or white-labeled third-party decisioning software, which slowed every implementation and pushed key logic outside the platform.
Rules studioBilling · Policies · Draftready for approval

Describe a policy in plain English

12 / 12 cases passed

Agent activity

    An illustration of the pattern, not my employer's code or data.

    Risk assessment workflows

    Based on the custom workflows for risk assessments from my time at OneTrust. Answer the intake questions and the workflow builds itself: risks get scored, stages are added or skipped by rule, and approval goes to the right person.

    workflow rulesrisk scoringapprovals
    Who it's for
    Privacy, security and risk teams, plus the business owners who request new vendors and systems
    Market
    Privacy, GRC and responsible AI software
    Problem
    Every company assesses risk its own way. Fixed templates pushed routing and approvals back into email and spreadsheets, so reviews stalled and risks lived outside the system.

    Intake · new AI vendor

    Workflow

      Risk heat map

      LikelihoodImpact

        Rules fired

          Score = likelihood × impact: 12 and up is high, 5 to 11 medium. A simplified illustration of the pattern, not OneTrust's product or data.

          Ecodrop Technologies

          The recycling startup I co-founded. Scan an item to compare the computer vision we shipped with a multimodal LLM rebuild, and watch the leaderboard move.

          computer visionReact NativeTechstars · AWSgamification
          Who it's for
          University students and staff, plus local merchants who fund the rewards
          Market
          University recycling and sustainability programs
          Problem
          People want to recycle but don't know what's accepted, and nothing rewards doing it right. Wrong items contaminate whole loads and cost programs money.
          Thousandsof users
          4university customers
          $8KMRR
          $200K+raised, Techstars and AWS
          Take a picture of your materials

          Trace

            Rewards · 140 points

            Vision labels and LLM outputs are examples. Reward tiers come from the original partner deck.

            Campus leaderboard · this week

              Live feed

                Recognito

                A side project: search a video for the moment a word is spoken, then jump to it. The original transcribed video with recurrent neural networks.

                speech alignmentsearchbrowser extension
                Who it's for
                Journalists, legal teams, researchers and everyday viewers of long video
                Market
                Media houses, bar associations, public-affairs archives and video platforms
                Problem
                Finding one moment in a long hearing, lecture or interview means dragging the time bar or digging through comments. Search should work on what was said.
                TRY

                  Transcript · click any word

                  Word timings come from aligning the captions to the audio. Clip: Elephants Dream (2006), Blender Foundation, CC BY 2.5.

                  How the first version worked

                  v0 · Chrome extension

                  Search a word within the video:
                  nothingSearch
                  Found 'nothing' at:
                  3:314:522:13

                  Recreated from the original popup.html and popup.js.

                  What was real

                  // video_control.js, the v0 content script
chrome.runtime.onMessage.addListener(
  function(stamp) {
    var video = document.querySelector("video");
    video.currentTime = stamp;
    video.pause();
  }
);

                  The popup and player control worked on any page with a video. Transcription ran in the web app; the extension's search results were placeholders that returned the same three timestamps for every query.

                  What this demo adds to the extension

                  Search
                  3 fixed timestamps every spoken word, indexed
                  Precision
                  to the second to about 0.1 s, per word
                  Queries
                  one word word, prefix or phrase
                  Results
                  a list marks on the timeline and a live transcript

                  Blueprint vision agent

                  Built for a residential construction business I support on technology and product. It reads an architect's drawing into an estimate-ready room schedule and flags anything it isn't sure about.

                  visiondocument extractionhuman review
                  Who it's for
                  Estimators and project managers at small residential builders
                  Market
                  Residential construction
                  Problem
                  Every estimate and material order starts with measuring each room off the drawings by hand. It is slow, and one misread dimension flows into the whole budget.
                  SAMPLE PLAN · MAIN FLOOR · 1/4" = 1'-0"

                  Agent trace

                    1,824
                    Heated sf
                    684
                    Unheated
                    2,508
                    Gross

                    Cost-control agent crew

                    Built for the same business. Text a cost and a crew of agents files it, checks for duplicates and tests the budget. Anything risky waits for a person.

                    multi-agentorchestrationhuman approval
                    Who it's for
                    Owners and field crews at small builders, who are not spreadsheet people
                    Market
                    Small construction businesses
                    Problem
                    Costs get paid on site and never reach the spreadsheet. The budget was always behind, so overruns showed up late.
                    9:41Sample project
                    Job Cost
                    Two-lot build · sample data

                    Budget used

                    Latest

                      Agent crew · runs on every entry

                      Try one

                      The shipped version turned a texted cost into a ledger row in one step. This shows the multi-agent version, with sample data.

                      Sheet 02 Method

                      How I build with agents

                      Anyone can call an AI model. The product work is deciding what it can see, how we know it's right and when a person signs off.

                      Start with one agent

                      Most workflows need one capable agent. I add more only when steps can run independently, like checking a payment for duplicates and budget at the same time. Fewer agents means lower cost and easier fixes.

                      Buy the plumbing

                      Cloud platforms now host agents, their memory and long-running tasks. I use that and spend the team's time on what customers notice: what the agent can access and when it needs sign-off.

                      Define "correct" first

                      Before building, we write the test cases that describe a good answer. Every release must pass them, and every mistake found later becomes a new test, so quality only goes up.

                      Feed it the right information

                      Most AI mistakes come from missing information, not bad instructions. I connect agents to the systems where the data lives and keep what they see focused.

                      Earn autonomy step by step

                      Agents start with a person approving each action. As results prove reliable, approvals relax. Anything that can't be undone, like moving money or deleting data, keeps a person in the loop.

                      Budget every task

                      Each task gets a cost and speed target. Routine steps go to smaller, cheaper models and judgment calls go to the strongest one. It's the unit economics I learned in finance.

                      Sheet 03 Journey

                      Finance, consulting, product

                      I started in finance, moved into technology consulting, then into product management. Co-founding a Techstars-backed startup taught me to build from zero, and I bring that to enterprise products.

                      201820192020202120222023202420252026
                      Enterprise
                      Founder

                      Range

                      Strategy, discovery and go-to-market

                      Technical depth, from architecture to data

                      Pricing and unit economics

                      Hands-on building, from apps to agents

                      AI credentials

                      AWS Certified AI Practitioner

                      Azure AI Fundamentals (AI-900)

                      Agentic AI (Andrew Ng)

                      Google Cloud Generative AI Leader

                      AI Product Management Certification, Product School

                      Also

                      U.S. Patent Pending 63/502,475

                      Certified Scrum Product Owner

                      Product Analytics, Product School

                      SAFe 5 Agile

                      Sheet 04 Contact

                      Let's compare notes.

                      I'm glad to talk product, platforms and where AI is heading. Email is the fastest way to reach me.

                      LinkedIn ↗ Resume (PDF)