Skip to content
ABAryan Bakshi

Vol. 20 — No. 5
Summer 2026

Shipped

$4.99
Free for recruiters

The Systems Issue · Real-time media ★ AI ★ Full-stack ★ Data ★ Low-level

Exclusive · SDE Intern @ AWS · Summer 2026

I build production-gradesoftware across real-time media, AI, full-stack products & systems.

CS @ Purdue. AWS SDE Intern. Builder of OpenSeat, Barbera, AI campus tools, data systems, and low-level systems projects.

View ProjectsView Resume
Aryan Bakshi in a navy suit, smiling in an office atrium — halftone print treatment

Aryan BakshiCS @ Purdue · SDE Intern @ AWS

Software engineer building production systems across real-time media, full-stack products, AI tools, data platforms, and low-level systems.

Breaking

86% FEWER ABANDONED CALLS AT AWS — 22X SPEEDUP AT ELI LILLY — 325+ STUDENTS ON OPENSEAT — 180+ USERS ON BARBERA — 1,039 TESTS IN LOB CITY — MAPE CUT 27.4% → 9.8% AT BASF — 59 SEGREGATED FREE LISTS — CUSTOM MALLOC — 10-DEVELOPER TEAM LED

01The Editor’s Letter

Ambiguous problems in, shipped systems out.

I’m a CS student at Purdue who treats every project like a product: real users, real stakeholders, measurable impact. That’s meant shipping pre-call device readiness for enterprise calling at AWS, building a campus platform Purdue Libraries actually operates on, running a solo marketplace with live payments, leading a 10-developer AI app team — and going low enough to write my own shell and memory allocator in C.

The theme of this issue — and of everything I build — is that practical systems beat polished demos.

// Currently

SDE Intern @ AWS — WebRTC device readiness for Amazon Connect web calling.

B.S. Computer Science @ Purdue (SWE & AI concentrations), math minor — May 2028.

Portrait of Aryan Bakshi

— Aryan Bakshi

Editor-in-chief, builder of this issue

House rules

01

Ship through ambiguity

Take a fuzzy problem — “students can’t find study space” — and turn it into a deployed system with real users and a feedback loop.

02

Full-surface engineering

Frontend, backend, APIs, databases, AI integrations, data pipelines, WebRTC and media, systems programming — work wherever the problem actually lives.

03

Product before code

Stakeholder interviews, design docs, tradeoff writeups, sprint planning, leading teams. Engineering is a product discipline, not just an implementation one.

04

Production is the bar

Real users, real clients, real deployments — including the side projects. If it isn’t shipped and being used, it isn’t done.

02The Feature Well

Projects with real users.

Not class assignments — systems in production with stakeholders, metrics, and users. Swipe, click the side spreads, or use the arrows.

FeatureAWS · SDE Intern Project

AWS logo

Amazon Connect Device Readiness

Pre-call WebRTC device readiness for Amazon Connect web calling.

Refactored the Chime SDK React component library to decouple dependencies — enabling pre-call device readiness with persistent media state through handoff, plus WebRTC diagnostics for UDP/TCP reach, STUN/TURN relay, and ICE candidate gathering.

86%
lower abandonment & handle time
~1.5 min
saved per agent call
4.5K+
enterprise customers
WebRTCTypeScriptReactChime SDKAmazon ConnectICE / STUN / TURN
Chime SDK LibraryAmazon Connect

PG. 01 / 06 — Amazon Connect Device Readiness

03The Career Column

Where I’ve shipped.

  1. AWS logo

    Amazon Web Services

    Software Development Engineer Intern

    NOWJun – Aug 2026Mountain View, CA
    • Refactored the amazon-chime-sdk-component-library-react architecture to decouple dependencies — enabling WebRTC pre-meeting device readiness with full backwards compatibility and persistent state through media handoff.
    • Built the pre-call device setup experience for Amazon Connect Customer in-app web calling — reducing contact abandonment and average handle time by 86% and optimizing agent calls by ~1.5 minutes for 4.5K+ enterprise customers.
    • Extended the Connect Endpoint Test Utility with WebRTC connectivity diagnostics: UDP/TCP reach, STUN/TURN relay, ICE candidate gathering, and background-filter asset fetching.
    WebRTCReactTypeScriptChime SDK
  2. Eli Lilly and Company logo

    Eli Lilly and Company

    Software Engineering Intern

    Jun – Aug 2024Indianapolis, IN
    • Built Python/REST API automation for Veeva Vault user-deletion workflows — a 22x processing speedup saving ~37 developer-hours annually.
    • Supported a compliance workflow relied on by 1,200+ Global Regulatory Affairs personnel.
    PythonREST APIsVeeva VaultAutomation
  3. BASF logo

    BASF

    Data Science Intern

    Aug 2024 – May 2025West Lafayette, IN
    • Built Python data pipelines and forecasting workflows over 20,000+ agricultural supply-and-use records.
    • Reduced MAPE from 27.4% to 9.8% across 50+ years of market data using ARIMA/LSTM and visualization tooling.
    PythonARIMALSTMETLVisualization
  4. PurdueTHINK Consulting logo

    PurdueTHINK Consulting

    Project Manager · Consultant

    • Led a 5-person consulting engagement for Purdue Libraries: shipped OpenSeat, implemented wayfinding and signage recommendations across 67 audited signs, and automated staff workflows.
    • Supported client partnership conversations with Occuspace and PMA as the project expanded.
    ProductConsultingNext.jsStakeholders

04Departments

Engineering range.

Depth in six directions — from browser media internals down to raw heap memory.

Real-time media

01

Pre-call readiness, media handoff, and connectivity diagnostics in the browser.

  • WebRTC
  • Chime SDK
  • ICE / TURN
  • Device permissions
  • Browser media APIs

Full-stack products

02

Typed end-to-end apps with real users, from schema design to deploy.

  • Next.js
  • React
  • TypeScript
  • Supabase
  • PostgreSQL
  • Vercel

AI applications

03

Server-side AI architecture that keeps keys safe and outputs structured.

  • Gemini
  • RAG
  • Prompt & context injection
  • Server-side AI architecture

Data systems

04

Pipelines and forecasting that turn messy records into decisions.

  • ETL
  • Forecasting
  • ARIMA / LSTM
  • Visualization
  • Cron pipelines

Payments & security

05

Money-moving flows built to be idempotent, verified, and auditable.

  • Stripe Connect
  • Webhooks
  • RLS
  • Auth
  • HMAC
  • API validation

Systems programming

06

Comfortable below the runtime — processes, descriptors, and raw memory.

  • C
  • malloc / free
  • Shell execution
  • Signals
  • Pipes
  • Parsing

05The Stats Page

By the numbers.

0%
drop in abandonment & handle time at AWS
0.0K+
enterprise customers on Connect flows
0
AWS design documents authored
0
architectural layers decoupled
0x
processing speedup at Eli Lilly
0
developer-hours saved annually
0+
regulatory personnel supported
0+
agricultural records cleaned
27.4% → 0.0%
forecast MAPE after ARIMA/LSTM
0+
visualizations delivered at BASF
0
signs audited for wayfinding
20–25s → ~0s
staff booking review, automated
0.0K
monthly data entries analyzed
0+
Barbera users in production
0
automated tests in LOB CITY
0
segregated free lists in custom malloc

All figures from shipped, production work — audited by the editor.

Skip to content
AB
Portrait of Aryan BakshiAryan BakshiSoftware Engineer · CS @ Purdue University

SDE Intern @ AWS · Summer 2026

I build production-grade software across real-time media, AI, full-stack products, and systems.

CS @ Purdue. AWS SDE Intern. Builder of OpenSeat, Barbera, AI campus tools, data systems, and low-level systems projects.

View ProjectsView Resume

Software engineer building production systems across real-time media, full-stack products, AI tools, data platforms, and low-level systems.

connect · device-readiness
PASS

  • Microphone

    granted

  • Camera

    1080p ready

  • Speaker

    output ok

  • Network

    TURN relay · 42ms

✓ ice: 6 candidates gathered

✓ webrtc: udp/tcp reachable

→ media state persisted through handoff

01Engineering identity

Ambiguous problems in, shipped systems out.

I’m a CS student at Purdue who treats every project like a product: real users, real stakeholders, measurable impact.

That’s meant shipping pre-call device readiness for enterprise calling at AWS, building a campus platform Purdue Libraries actually operates on, running a solo marketplace with live payments, leading a 10-developer AI app team — and going low enough to write my own shell and memory allocator in C.

Currently: SDE Intern @ AWS, working on WebRTC device readiness for Amazon Connect web calling.
B.S. Computer Science @ Purdue, math minor — concentrations in Software Engineering & AI, May 2028.

aryan@purdue:~$ whoami
Aryan Bakshi in a navy suit, smiling in an office atrium
01

Ship through ambiguity

I like taking a fuzzy problem — “students can’t find study space” — and turning it into a deployed system with real users and a feedback loop.

02

Full-surface engineering

Frontend, backend, APIs, databases, AI integrations, data pipelines, WebRTC and media, systems programming — I work wherever the problem actually lives.

03

Product before code

Stakeholder interviews, technical design docs, tradeoff writeups, sprint planning, and leading teams. Engineering is a product discipline, not just an implementation one.

04

Production is the bar

Real users, real clients, real deployments — including the side projects. If it isn’t shipped and being used, it isn’t done.

02Featured work

Projects with real users.

Not class assignments — systems in production with stakeholders, metrics, and users. Swipe, click the side cards, or use the arrows.

AWS · SDE Intern Project

AWS logo

Amazon Connect Device Readiness

Pre-call WebRTC device readiness for Amazon Connect web calling.

Refactored the Chime SDK React component library to decouple dependencies — enabling pre-call device readiness with persistent media state through handoff, plus WebRTC diagnostics for UDP/TCP reach, STUN/TURN relay, and ICE candidate gathering.

86%
lower abandonment & handle time
~1.5 min
saved per agent call
4.5K+
enterprise customers
WebRTCTypeScriptReactChime SDKAmazon ConnectICE / STUN / TURN
Chime SDK LibraryAmazon Connect

01 / 06 — Amazon Connect Device Readiness

03Experience

Where I’ve shipped.

  1. AWS logo

    Amazon Web Services

    Software Development Engineer Intern

    NOWJun – Aug 2026Mountain View, CA
    • Refactored the amazon-chime-sdk-component-library-react architecture to decouple dependencies — enabling WebRTC pre-meeting device readiness with full backwards compatibility and persistent state through media handoff.
    • Built the pre-call device setup experience for Amazon Connect Customer in-app web calling — reducing contact abandonment and average handle time by 86% and optimizing agent calls by ~1.5 minutes for 4.5K+ enterprise customers.
    • Extended the Connect Endpoint Test Utility with WebRTC connectivity diagnostics: UDP/TCP reach, STUN/TURN relay, ICE candidate gathering, and background-filter asset fetching.
    WebRTCReactTypeScriptChime SDK
  2. Eli Lilly and Company logo

    Eli Lilly and Company

    Software Engineering Intern

    Jun – Aug 2024Indianapolis, IN
    • Built Python/REST API automation for Veeva Vault user-deletion workflows — a 22x processing speedup saving ~37 developer-hours annually.
    • Supported a compliance workflow relied on by 1,200+ Global Regulatory Affairs personnel.
    PythonREST APIsVeeva VaultAutomation
  3. BASF logo

    BASF

    Data Science Intern

    Aug 2024 – May 2025West Lafayette, IN
    • Built Python data pipelines and forecasting workflows over 20,000+ agricultural supply-and-use records.
    • Reduced MAPE from 27.4% to 9.8% across 50+ years of market data using ARIMA/LSTM and visualization tooling.
    PythonARIMALSTMETLVisualization
  4. PurdueTHINK Consulting logo

    PurdueTHINK Consulting

    Project Manager · Consultant

    • Led a 5-person consulting engagement for Purdue Libraries: shipped OpenSeat, implemented wayfinding and signage recommendations across 67 audited signs, and automated staff workflows.
    • Supported client partnership conversations with Occuspace and PMA as the project expanded.
    ProductConsultingNext.jsStakeholders

04What I build

Engineering range.

Depth in six directions — from browser media internals down to raw heap memory.

Real-time media

Pre-call readiness, media handoff, and connectivity diagnostics in the browser.

WebRTCChime SDKICE / TURNDevice permissionsBrowser media APIs

Full-stack products

Typed end-to-end apps with real users, from schema design to deploy.

Next.jsReactTypeScriptSupabasePostgreSQLVercel

AI applications

Server-side AI architecture that keeps keys safe and outputs structured.

GeminiRAGPrompt & context injectionServer-side AI architecture

Data systems

Pipelines and forecasting that turn messy records into decisions.

ETLForecastingARIMA / LSTMVisualizationCron pipelines

Payments & security

Money-moving flows built to be idempotent, verified, and auditable.

Stripe ConnectWebhooksRLSAuthHMACAPI validation

Systems programming

Comfortable below the runtime — processes, descriptors, and raw memory.

Cmalloc / freeShell executionSignalsPipesParsing

05Impact

Numbers that shipped.

0%
drop in abandonment & handle time at AWS
0.0K+
enterprise customers on Connect flows
0
AWS design documents authored
0
architectural layers decoupled
0x
processing speedup at Eli Lilly
0
developer-hours saved annually
0+
regulatory personnel supported
0+
agricultural records cleaned
27.4% → 0.0%
forecast MAPE after ARIMA/LSTM
0+
visualizations delivered at BASF
0
signs audited for wayfinding
20–25s → ~0s
staff booking review, automated
0.0K
monthly data entries analyzed
0+
Barbera users in production
0
automated tests in LOB CITY
0
segregated free lists in custom malloc