Muhammad Izaan Shaikh — Systems & AI Product Engineer

Case 01 — Flagship

Savvanta — hiring,run as an engineered system

A hiring panel deliberating around a table

Fig. 01 — the panel, the gut feel, the guesswork

Problem · unverifiable skill

CVs are claims. Interviews are impressions. Neither is evidence.

Screening is slow, biased and easy to game — and generated code has made unproctored take-homes worthless. Employers needed a way to watch real skill happen, not read about it.

SAVVANTA — SYSTEM TRACE · 5 REPOSITORIESLIVE

ClientVue + embedded Monaco IDE for in-browser coding

API CoreNestJS — every domain rule enforced server-side

LLM ServiceFastAPI — AI-generated challenges, AI-graded attempts

SandboxgVisor isolation, Nix-provisioned toolchains

Token BrokerPer-challenge AI budgets, enforced in real time

candidate session — gVisor cell #4021

Inside the sandbox · repo 04

30+ languages, spun up per assessment.

Untrusted candidate code runs inside gVisor cells; Nix shelves every toolchain and provisions exactly the languages each assessment needs. Scroll — this is the provisioner warming up.

RUNTIME SHELF0/30 PROVISIONED

The governor · repo 05

AI is metered like electricity.

Every challenge gets a token budget. The broker meters each candidate's AI usage against it and cuts over-consumption before it bills — watch it enforce.

Live productsavvanta.comRanked talent dashboards

Challenge #88 · Token budget0 / 50,000 tok

Case 02 — Medical Imaging

Oncosoft — segmentationat software prices

A radiologist reading scan films on a lightbox

Fig. 02 — expert hours, spent drawing outlines

Problem · the contouring bottleneck

Radiotherapy can't start until someone draws the tumour — by hand.

Contouring a single patient's CT set consumes expert hours, and the commercial autosegmentation suites that automate it are licensed at prices most hospitals can't justify. Oncosoft trains its own models — on public datasets and hospital-partnered, privacy-scrubbed data — and delivers the full pipeline as affordable SaaS: training, secure storage and retrieval, clinical workflow.

AUTOSEGMENTATION · SLICE 001/128

Training pipelineSecure data storage & retrievalHospital-partnered training dataFull commercial SaaS

Case 03 — Research Statistics

Kyrostat — analysis withoutthe waiting room

Fig. 03 — the analysis, weeks after the question

Problem · the biostatistician queue

Every research question waits weeks in someone else's queue.

Medical researchers budget for a biostatistician, wait days or weeks for results, and get a limited number of revisions. Kyrostat reads your data and research context, suggests the right tests and parameters, explains the assumptions behind every choice — and lets you iterate infinitely until the analysis is yours.

SUGGESTED TEST · loading…

Test & parameter suggestionsAssumptions, explainedUnlimited iterationkyrostat.com — live

Case 04 — Guidance at Scale

Askedii — ten questions,a mapped future

Fig. 04 — a life decision, made on hearsay

Problem · choosing blind

Choosing a university is a life decision made almost blind.

Most school-leavers pick a degree on hearsay — no idea which universities fit, what scholarships exist, or what a campus is actually like. Askedii asks ten questions, matches ten universities, opens a per-university chatbot for the details — and connects applicants to paid mentor calls with students already there.

MATCHING · 10 QUESTIONS → 10 UNIVERSITIES

Working With Me

Fixed prices. Working software.No hourly meter.

Phase 01

Discovery sprint

One to two weeks. I map your problem into an architecture, a technical spec, and a fixed-price phased plan you could hand to any team.

Paid — it filters both ways

Phase 02

Build, in phases

Each phase is fixed-scope, fixed-price, and ends with working software you can run — not a progress report.

You see it working every phase

Phase 03

Handover

Deployment, documentation your future team can build on, and a 30-day post-launch support window.

It's yours, fully

Then choose how you own it

One-time

Ownership

Buy it outright

One fixed price for the whole build. At handover, everything is yours — no strings, no meter.

Full source code and infrastructure, transferred to you
Documentation your future team can build on
30-day post-launch support window
Zero ongoing obligation

One payment · Full handover at Phase 03

Monthly

Managed

I run it for you

A monthly plan — I host, maintain and keep improving the software while you run the business.

Hosting, monitoring and security patching handled
Priority fixes with agreed response times
Continuous improvements shipped every month
Monthly health and usage report

Lower entry · Switch to full ownership anytime

Bring me the project other agencies said was too hard.

One conversation is enough to know whether I'm the right person to build it. If I'm not, I'll tell you.

[email protected]

GitHub LinkedIn Savvanta Kyrostat