Open to Senior Full-Stack & AI Engineering roles · Bangalore / Remote

Phaneendra Bheesetti

Senior Full-Stack & AI Engineer (6+ years) shipping LLM systems that hold up in production — a real-time voice agent, LLM eval pipelines, sandboxed code execution — now leading AI initiatives at BMW, after architecting enterprise platforms used by 1,500+ people at Airbus.

TypeScript-first, with production depth in Angular, React, Node.js, AWS, and real-time AI systems. Founder of KodeRyu, a live AI voice interviewer. IIT Madras alum. Bangalore · Open to remote.

  • €1.5B+
    cumulative budget governed
    ABAcS
  • 1,500
    users across 5 countries
    ABAcS
  • 70%
    faster allocation cycles
    ABAcS
  • ≈4,700 hrs
    specialist effort eliminated
    ECM Tool
01 Selected Work

Systems I owned end-to-end

Four builds across aviation and AI — architecture-level detail, real numbers, shipped to real users.

ABAcS Budgeting Tool

Airbus · Sole architect

Airbus finance teams tracked a €500M+ annual budget across five countries in fragile, disconnected spreadsheets.

Solo-architected a full-stack platform governing €500M+/yr for 1,500 users across France, Germany, UK, Spain and India — 70% faster allocation cycles, audit prep days to hours — work conventionally staffed by a 3–5 engineer team, delivered solo.

  • Governs €500M+ of budget allocation per year — €1.5B+ cumulative over 3+ years live — with regulatory-grade, multi-jurisdiction controls.
  • Multi-tiered role-based access adaptable to German, EU and Indian financial regulations, EU-guideline encryption, dashboards tailored to 6 organizational levels.
  • Integrated Google Sheets as the sanctioned live data layer with a caching layer and global search; interviewed users at every level and shipped a feedback-telemetry loop.
ReactJavaScriptMaterial UIGoogle Apps ScriptSheets APICrypto JS

ECM Tool

Airbus · Desktop app

Publishing classified documents into the central archive relied on a brittle legacy Excel-macro workflow that took hours per file.

Architected an Electron desktop app end-to-end — 2–3 hrs down to 10–15 min per document (≈90%) — across 2,000+ documents to date, ≈4,700 specialist hours eliminated (≈2.4 person-years); one publisher now does a manual team's work.

  • Implemented OLE container extraction and a PDF processing pipeline over complex embedded-document structures.
  • Integrated a CFB library to parse binary compound-file containers inside Word files — deep-systems debugging most workflows never touch.
  • Built a modular plugin for signature removal, metadata extraction and batch processing — composable publishing workflows.
ElectronReactNode.jsAdmZipPDF-lib

MPO (Skywise)

Airbus · Reliability focal

Airbus's maintenance-program optimization platform (Palantir Foundry) tunes each maintenance task's interval — increase, decrease, keep or delete — across single-operator, multi-operator and fleet-wide exercises, carrying a multi-million-euro annual cost-avoidance and service-revenue business case.

Owned the consumption layer — the digitally-signed PDF dossier that is the revenue-carrying deliverable operators receive — and resolved 15+ critical production defects on a revenue product, as the team's incident-review focal across 8,000+ operator accounts.

  • Built the dossier pipeline via a headless-Chromium renderer with embedded digital signatures — the signed PDFs that deliver each optimized maintenance program to operators.
  • Hardened reliability with Jest unit and integration coverage; owned incident review on a live revenue product.
  • Shipped increments and fixes in close partnership with business stakeholders, speeding customer-issue turnaround.
Node.jsReactReduxPalantirAWSJenkinsJest
Currently Building

Founder · Live at koderyu.com

KodeRyu ↗

Interview-prep tools grade code but never rehearse the live, spoken back-and-forth of a real technical interview.

Built solo end-to-end — an AI voice interviewer that conducts real spoken technical interviews across voice, DSA and system design, built on a real-time streaming voice loop and a sandboxed code executor — live at koderyu.com, heading to public launch.

  • Voice agent that holds a natural spoken interview in real time — streaming speech-to-text (Deepgram) into an LLM and back through self-hosted TTS (chosen for latency and unit cost) fast enough to feel like live conversation, recovering when candidates interrupt and hinting progressively when they're stuck.
  • Executes untrusted candidate code in 5 languages inside a Docker + nsjail sandbox on a BullMQ worker pool, with sub-second cold starts.
  • LLM-as-judge grading that matches human expert graders — κ=0.92 agreement on a 13-criterion rubric — regression-tested by simulating 12 distinct candidate types.
React 18TypeScriptExpressPostgresDockernsjailBullMQ
02 Projects

Also shipped

Quality Tracker Tool

Airbus · Solo build, 2 months

Full-stack supply-chain quality tracker delivered to 500+ users across 5 countries. Cut the approval cycle from 1–2 days to 2–3 hours — same-day resolution of issues that used to leave production exposed for days — with zero bug-induced downtime, then handed to a 2–3 engineer team that extended it.

Composite Compatibility Tool

Airbus · 2-engineer build

Web app automating compatibility detection for aircraft composite materials — global search, ag-Grid tables, Firebase auth and a Postgres schema. Cut testing cycle time by 80%.

MyMetrics

Société Générale · Finance Risk

Raised a 60K-line PnL & risk platform from a SonarQube D to A+ over 10 months, and built 5 new APIs powering risk and PnL dashboards for SG business lines.

03 About

Six years of shipping things people depend on

I'm a Senior Full-Stack & AI Engineer with 6+ years building production-grade systems across automotive, aviation, finance and healthcare. Today I work with BMW's engineering division (via Alten), leading full-stack delivery on a large engineering-data platform — and I own the team's AI initiatives: designing agent-based development workflows, building MCP servers that connect internal systems to AI tooling, and shaping what AI-assisted engineering looks like for the team — with the same eval-first discipline I apply in my own AI products.

Before that, six years at Airbus turning messy business requirements into tools that teams in five countries rely on every day — I like being the person who owns a system from the first architecture sketch to the on-call rotation. I studied at IIT Madras, with a dual degree in Civil Engineering and a minor in Operations Research.

Nights and weekends go into KodeRyu, an AI voice interviewer I'm solo-founding — live at koderyu.com and heading to public launch. I'm happiest at the intersection of deep technical work and sharp product instinct — talking to the actual users, then shipping the thing that makes their day shorter.

  • BasedBangalore, India · open to remote
  • NowSenior Full-Stack & AI Initiatives @ BMW (via Alten) — agents, MCP servers, AI dev workflows
  • FocusAI engineering — LLM apps, agents, evals · Full-stack — TypeScript, Angular, React, Node.js, AWS
  • DomainsAutomotive · Aviation · Finance · Healthcare
  • EducationIIT Madras — Dual Degree, Civil Eng. (Minor: Operations Research)
  • FounderKodeRyu — AI voice interviewer · live at koderyu.com