I'm a BS Computer Science student at Bahria University (Islamabad), focused on full-stack development and applied AI. I learn by shipping — most of my time goes into building real apps end-to-end.

AI/ML Engineer Intern @ Nerd Flow (Jul 2026 — Sep 2026) — building the AI hiring stack for Preplify, their HR management platform:

• LLM résumé parsing that turns raw CV text into structured candidate profiles
• Per-criterion candidate screening against the job description (skills, experience relevance, education fit) — with HR able to override every automated decision
• AI-conducted first-round interviews, transcribed and scored the same way
• Benchmarked self-hosted models (Qwen2.5-1.5B via Ollama, fine-tuned Qwen3-0.6B) against hosted OpenAI APIs on accuracy, cost-per-résumé, and keeping candidate data in-house

• AI / ML: LLM APIs (OpenAI), prompt design, Ollama / local models, pandas, NumPy, scikit-learn
• Mobile: Flutter, Dart, Riverpod
• Backend: Node.js, Express, PostgreSQL
• Frontend: React, Vite, Tailwind, HTML/CSS/JS
• Languages: C++, Python, JavaScript, Dart
• Cloud: Firebase, Render, Vercel, Neon

• Qeemat (formerly Hisaab — private repo) — A free, offline-first personal-finance and receipt manager for Pakistan. Flutter + Node/Postgres + React admin, built solo, deployed entirely on free tiers at zero monthly cost.

  • Automatic transaction capture — reads bank and wallet alerts through an Android notification listener and parses them entirely on-device; the message text never leaves the phone, which is enforced by tests rather than only promised. Low-confidence captures wait for a one-tap confirmation instead of silently posting.
  • Categorization engine — four layers (a rule you wrote, a rule it learned, a lexicon covering English, Urdu and Roman Urdu, then amount-and-timing as a tie-break). Every entry shows where its category came from, and anything it is unsure about goes to a review queue rather than being filed on a guess.
  • Bill splitting — equally, by amount or by percentage, with the parts always reconciling to the bill to the paisa. Splits open lent-direction loans rather than a second ledger, so "you are owed" has one number that cannot disagree with itself.
  • Security — biometric app lock with PIN fallback and attempt lockout, screenshot blocking, a Hive cache encrypted with a key in the Android Keystore, 15-minute access tokens with rotating refresh tokens and replay detection that revokes the whole family. Receipt images are private assets reached only by signed URL, with EXIF stripped before upload so a photo taken at home does not carry the address.
  • Roles that hold — moderators can see aggregate analytics and nothing that identifies a user; every admin access to an individual's records is written to an audit log. Enforced at the API and covered by a suite that probes all 41 id-taking routes with another user's identifiers.
  • Real money model — multi-account balances, transfers that are never counted as spending, udhaar (lent/borrowed), household members with allowances, hierarchical categories, monthly reports, CSV export and a printable PDF statement typeset on the device.
  • Offline outbox — ordered, idempotent replay keyed on client-generated ids, verified in airplane mode. 107 backend tests in CI against a real Postgres, plus 66 on the Flutter side.

• CineWorld — Full-stack cinema ticketing and management system: 9-table normalised MySQL schema, role-guarded admin dashboard, printable PDF tickets.

• CL Compiler — JavaCC/JJTree compiler front end plus a JavaFX playground for the Classroom Language.

• FoodBook MVP (team project) — Shipped the MVP of a community food review platform with another developer: REST APIs, unit tests, beta bug-fixing.

• Portfolio: sameer-hassan01.github.io
• CV: sameer-hassan01.github.io/Sameer_Hassan_CV.pdf
• Email: sameerhassanq3@gmail.com

Open to internships and junior software engineering roles.