JoVE · Product Manager
AI answers JoVE can prove.
My first of two bets for JoVE - live, right here. It grounds every answer in a peer-reviewed protocol and cites it, or admits it can't.
In the transfer step, the size-separated proteins are moved out of the gel onto a solid-phase membrane, assembled into a transfer sandwich under set transfer conditions, ready for antibody-based detection of the target protein P1P2.
- BM25 retrieval over real JoVE protocols
- Live LLM answer, grounded in retrieved passages
- Per-claim citations to peer-reviewed video
- Scorecard runs the same pipeline (falsifiable)
- Rate-limited API route
- Corpus is a 9-protocol slice of JoVE's 16K+
- Impact dashboard uses sample data
- Revenue projections are indexed & directional
Independent study · not affiliated with JoVE · JoVE, competitor & market facts from public sources (2024-26)
The scorecard, side by side.
Four method questions, run through the same retrieval and model you just drove above.
I can't ground this in JoVE's peer-reviewed protocols, so I won't guess. The cited T-cell CRISPR method covers the workflow, but not this specific concentration.
The corpus covers the workflow but not this exact number - so Verify says so instead of inventing one.
Use about 1 µM gRNA complexed 1:1 with Cas9 for a CRISPR knockout in T cells - that's the standard working concentration.
Why it's risky. Invents an exact concentration the corpus never states. This is the hallucination that ends up in a lab notebook.
Illustrative, pre-run comparison of the same two engines you can drive live above. Verify-side citations point to real JoVE protocols.
“Why pay for JoVE when AI is free?” The answer is the citation.
Kaushal Khodifad · Senior PM, OTPless · Founder, CLOZOM · IIM Ahmedabad
What else I'd build for JoVE.
You've just used the first bet. Here's the rest - and the 3-year plan.
Reading JoVE
What I see at JoVE.
The business compounds on these two levers. My two bets defend exactly those.
Proposal 02 · Defend the renewal
JoVE Impact: make the value impossible to miss.
Make the value self-serve: usage, citations, and cost per use, with a one-click ROI & renewal report.
JoVE records usage, but a librarian has to ask for it - so at renewal, the value nobody can see is the line item that gets cut.
Usage, citations, and cost per use, self-serve - with an inspectable ROI model the librarian can flex to their own numbers.
Product strategy · the 3-year bet
Where JoVE wins next.
Beyond the two proposals - the market, the roadmap, and the revenue it unlocks.
The white space. Only JoVE owns all three - peer-reviewed rigour, the video format, and an AI that already answers. Verify adds the missing discipline to take the open top-right corner.
Biotech & pharma already subscribe. Sell them audit-ready, cited video training.
| Capability | Generic AI ChatGPT et al. | JoVE today AI Search + Co-Pilot | JoVE + Verify the missing discipline |
|---|---|---|---|
| Answers a method question | |||
| Grounded in peer-reviewed video | |||
| Per-claim citation to a specific protocol | |||
| Abstains when it can't ground the answer | |||
| Shows the retrieval receipt (what it read) |
Public market reports & pharma R&D estimates, 2024-26.
And proof I can ship it.
Two products I led at CLOZOM - the contribution, and the numbers.
- CLOZOM foundedJun 2023India's healthcare professional network
- Built five verticals0 → 1four B2C + one B2B, owned end to end
- 50,000+ membersTodaylive nationwide across India
Built CLOZOM 0→1 to 50,000+ healthcare professionals across five verticals.
Founder · Jun 2023-present · India's healthcare professional network.
My contribution
- Owned product 0→1 - strategy, roadmap and all five verticals (4 B2C, 1 B2B).
- Scaled to 50,000+ registered healthcare professionals and 26,000+ live jobs nationwide.
- Set the north-star metrics and growth loops across acquisition, activation, retention.
Key results
Shipped a clinical AI assistant clinicians could trust.
Founder · CLOZOM · Clinical AI assistant inside CLOZOM.
My contribution
- Grounded every answer in vetted sources - it abstains rather than guess.
- Evaluated vendor guardrail options, set safety thresholds, benchmarked accuracy on an eval set.
- Same grounding + guardrails discipline behind Proposal 01 (Verify).