# Arslan Jan Soomro

**Senior Full-Stack AI Engineer** · Islamabad, Pakistan · Open to new roles · Remote

I build LLM products end to end, from RAG and agents to evals, and ship them to paying users. I was tech lead on the Pixis × Google co-development initiative and led a 6-engineer team at Pixis. On Claire IT AI, a medical-coding assistant, I raised retrieval hit rate from 28% to 75% and cut p90 latency by 40%.

- Email: arslanjs.dev@gmail.com
- CV (PDF): https://arslanjs.dev/cv/arslan-jan-soomro-cv.pdf
- LinkedIn: https://www.linkedin.com/in/arslanjs
- GitHub: https://github.com/Arslan-Soomro

## Scorecard

| Result | Baseline | Now | Change |
| --- | --- | --- | --- |
| Retrieval hit rate on pasted clinical charts (Claire IT AI) | 28% | 75% | +47 pt |
| p90 answer latency, validated on an eval set (Claire IT AI) | 83 s | 50 s | −40% |
| Async candidate screening, per 2 hours (Proctor) | 4 | 12 | 3× |
| Pull requests reviewed and merged, leading 6 engineers (AdRoom · Pixis) | – | ~170 | in 4 months |

## Selected work

- [Google Creative Coverage](https://arslanjs.dev/work/google-creative-coverage.md): I designed AdRoom’s side of the Pixis × Google co-development initiative, now live as Google Creative Coverage, and led my team to ship it to production. Role: Technical lead and solution architect. from my prototype to production, within two months: Prototype → Live.
- [Claire IT AI](https://arslanjs.dev/work/claire-it-ai.md): A medical-coding AI assistant. I raised its retrieval hit rate from 28% to 75% and cut p90 latency by 40%. Role: Lead engineer. failed chart searches, out of 20 eval charts: 10 → 0.
- [ai-ad-gen](https://arslanjs.dev/work/ai-ad-gen.md): An agentic ad generator that keeps product details intact, used in production by a Pixis enterprise client. Role: Creator and lead developer. with a Pixis enterprise client: In production.
- [Hammurabi](https://arslanjs.dev/work/hammurabi.md): An AI legal-analysis SaaS for Israeli lawyers, in Hebrew and English. I took it from first commit to production in about 3 months, then grew it into a paid product. Role: Founding & lead engineer. legal AI tools shipped on one framework: ~12.
- [AdRoom](https://arslanjs.dev/work/adroom.md): I took ownership of Pixis's AI creative platform from a previous agency and led its 6-engineer team, shipping UGC video, video scoring and credit billing. Role: Engineering lead. known critical and high security advisories on the pixis.ai CMS: 28 → 0.

## Experience

- **Apr – Oct 2026**, Engineering Lead, Pixis AI. Owned engineering for AdRoom, ai-ad-gen and pixis.ai. Led a 6-engineer team, reviewed and merged ~170 PRs in 4 months, and was tech lead on the Pixis × Google co-development initiative.
- **2025 – 2026**, Lead Engineer, Claire IT AI. Built the RAG orchestrator over ~169k medical codes. Raised retrieval hit rate from 28% to 75% and cut p90 latency by 40%, validated on an eval set.
- **2024 – 2026**, Founding & Lead Engineer, Hammurabi. Took a Hebrew and English legal-AI SaaS from first commit to production in about 3 months, then grew it into a paid product. Solo for 10 months, then led the team.
- **Jun 2024 – Oct 2026**, Co-Founder & Lead Engineer, Digitenic (now Revonix). AI engineering company building LLM products for clients in ad tech, legal tech, healthcare and fintech. Led engineering across the products above, plus BudgetHub, Pixis Visibility and Proctor, where async candidate screening went from 4 to 12 per 2 hours.
- **Jun 2023 – Aug 2024**, Founder, Automychat & Othersify. Built a multi-tenant WhatsApp AI-chatbot platform for small businesses, and led a 5-person startup that shipped a Gmail read-receipt extension.
- **Jun – Aug 2023**, Full Stack Developer, Dynalytix AI. Top frontend contributor on Doclytix, the company’s AI document-management SaaS: 292 of its 593 frontend commits and 37 merged PRs in 8 weeks.
- **Jun 2022 – Jul 2023**, Full Stack Developer, LeaseMagnets (Tour). #3 contributor of 43 on Tour’s React SaaS dashboard. Led its Next.js 13 rewrite and built billing, contracts and live co-browsing.
