bengaluru, in / --:-- ist
Hey, I'm Prateek. I studied aerospace engineering at UC San Diego, spent a few years building lab hardware at Illumina in San Diego, and these days I work on AI security products in Bengaluru. When I'm not at work you'll probably find me surfing (on a board I shaped myself), hanging around ropes courses, tinkering with something in a workshop, or explaining something nerdy to whoever will listen.
scrollthe work![[describe photo 1]](/assets/photos/01.jpg)
![[describe photo 2]](/assets/photos/02.jpg)
![[describe photo 3]](/assets/photos/03.jpg)
![[describe photo 4]](/assets/photos/04.jpg)
![[describe photo 5]](/assets/photos/05.jpg)
![[describe photo 6]](/assets/photos/06.jpg)
01 — Products
AI security products I've helped build at Globals — from testing models to guarding them in production.
AI red-teaming product
Adversarial evaluation for LLM applications — probing models for jailbreaks, prompt injection, data leakage and unsafe output, and turning results into reports security and engineering teams can act on.
Input-guard extension
A guardrail layer that screens prompts before they reach the model — catching injection attempts and policy violations at the point of input, without rebuilding the app.
Complete platform
Bringing testing, guarding and reporting into one platform so teams can run the full AI security lifecycle in one place instead of stitching tools together.
02 — Go-to-market
Presales across VAPT, GRC, cybersecurity services and government & deep-tech tenders — for an audience of C-suite, engineering directors and engineers.
End-to-end ownership from first discovery call to signed deal: 80+ discovery calls, 60+ proposals, 50+ demos in the defence ecosystem and 30+ proofs of concept.
Primary author or primary contributor on eight defence tender responses — translating technical capability into compliant, evaluable bids.
Participated in the IndiaAI Impact Summit and co-hosted a pre-summit event with BEL and Globals on AI security.
India channel-partner liaison for Vanta, and support on SOC 2, ISO 27001 and CMMC compliance engagements for clients.
03 — Writing
Published Anvil evaluations of how real LLMs hold up under adversarial pressure.
04 — Engineering
Process engineering at Illumina and hardware builds from UC San Diego.
Process optimisation on the flagship glass-slide consumable for the NovaSeq X.
9% unit cost cut · ~$600K est. annual savings
Python GUI that pulled from JIRA and internal databases to auto-notify clients about limited storage.
15% less storage needed · 10% less scrap
IQ/OQ/PQ validation reporting and data-driven KPI reports under ISO 13485:2016.
4% fewer defects on R&D orders
Led the mechanical build of a large-scale animatronic with servo drivers and IR sensors, run remotely through the pandemic.
30% under allocated budget
Designed in BoardCAD, CNC-cut and hand-shaped, then tuned through miniature wind-tunnel testing before glassing.
BoardCAD · CNC · epoxy & carbon fibre
Design analysis with GD&T and DFM on a case-hardened wing actuator in active development.
12% lighter · 8% cheaper per unit
05 — Experience
AI security & LLM evaluation SaaS. Product direction on the Anvil suite, plus presales and GRC engagements.
Helped ship Anvil and Anvil Shield; ran discovery, demos, PoCs and tender responses.
Cost and quality improvements on NovaSeq X consumables; liaison between development and internal customers.
Ran high-ropes and team-building programmes for school and corporate groups, end to end.
Led a student team building a synchronised animatronic.
Minor in International Business. GPA 3.8.
06 — Contact
Open to conversations about AI security products, presales and product roles.