Resume
Shlok Sharma
Product Engineer · AI Products · United States
shlokzartsharma@gmail.com ·
shloksharma.com ·
GitHub ·
LinkedIn
Summary
I ship LLM products end-to-end: product sense, systems design, evals, and the unglamorous ops that make something charge a card. Looking for founding or product-engineering roles at 5–100 person startups where real-user outcomes are the job.
Experience
Founder & AI Engineer — FlyingMinds.org
- Rebuilt the AI tutoring engine (Node, DeepSeek, Bayesian Knowledge Tracing) as sole engineer; scaled to 1,500+ students and $15K+ revenue.
- Shipped adaptive testing, gamified vocabulary, Stripe subscriptions, and an admin dashboard used to run the business day-to-day.
- Built an LLM grading eval harness that raised key agreement from 81% → 92%, then wired those gains back into production rubrics.
- Chose model routing and unit economics deliberately so a $29/month family plan could stay interactive without silent quality collapse.
Full-Stack Engineer — MaxedS
- Rebuilt the CRM interface and internal sales tools used daily by a fractional sales team — fewer clicks, clearer pipeline state.
- Prototyped an AI document-analysis tool and automation workflows that cut manual sales-ops work by ~40%.
- Prioritized workflows that removed handoffs over features that only looked impressive in demos.
Contract Developer — ibidPREP
- Rebuilding the marketing site for a NYC test-prep company on a milestone contract — front-end structure, visual design, conversion-minded pages.
Selected products
- IvyOdds — college admission probability product; web + native iOS; 1,000+ downloads and rapid early web growth.
- Driver’s Voice — Next.js organizing site with Claude extraction pipeline, Zod schemas, WhatsApp share kit, and OG image cards.
Education
Wesleyan University — B.A., Computer Science
How I work (not a laundry list)
TypeScript / Node when shipping product fast; Python when the eval or data path is clearer there; React/Next on the client; Postgres/Supabase when I need durable state; Stripe when money is real. I care more about eval loops, mastery/state modeling, and operational clarity than collecting framework logos.