Sharayu Rasal Software engineer
Reliable systems.From retrieval to approval.
I build the engineering around the model: retrieval that cites its sources, approval before anything commits, and the evaluation that shows how well it actually works.
Looking for full-time roles in AI engineering, backend, and cloud systems.
NYU M.S. Computer Science New York Los Angeles
Education.
Work that changed how I build.
Four roles across AI, data systems, computer vision, and product work. What changed, and what changed it.
- 01
Diyo
Jul 2026 – PresentAI Engineering & Data Intern
70%
less manual catalog review
80%
top-1 accuracy, from 30%
56%
fewer no-result scans
25%
lower cost per scan
- Cut manual catalog review by 70% by building an event-driven pipeline on EventBridge, Lambda, S3, and Glue that ingested and validated 800+ products across 16 categories.
- Raised top-1 product identification from 30% to 80%, reduced no-result scans by 56%, and lowered inference cost per scan by 25% through redesigned recognition and ranking, embedding caching, and batched inference.
- AWS
- Event-driven pipelines
- Computer vision
- ML inference
- 02
Bharat Forge
Jul 2023 – Nov 2023Software Research & Development Intern · Pune, India
18%
lower end-to-end latency
- Reduced end-to-end latency by 18% under constrained compute and network conditions by leading backend development of a real-time computer vision pipeline in C++ and Python.
- Stabilized throughput over unstable-bandwidth networks by designing reliable data transmission and buffering for the distributed capture-to-inference path.
- Collaborated with DRDO engineers to integrate embedded modules, documenting API specifications and hardware interfaces for production deployment.
- C++
- Python
- Computer vision
- Distributed data
- 03
Oniv Beverages
Oct 2023 – Feb 2024Technical Product Intern, Web & Marketing · Mumbai, India
30%
more organic reach
- Constructed and deployed a customer-facing marketing website, integrating Google Search Console to increase organic reach by 30%.
- Initiated a strategic partnership with Indobevs Brewery, leading to the #UnitedWeDrink x #TheFirstWinemakersOfIndia event and expanding brand engagement.
- Worked across web, growth, and product to run experiments, instrument analytics, and translate campaign results into product insights.
- Web development
- Analytics
- Product
- Growth experiments
- 04
Posit Source
Mar 2023 – Jun 2023Software Engineering Intern · Pune, India
60%
less manual reporting effort
30%
shorter lead time at peak
- Reduced manual reporting effort by 60% by shipping scalable Python backend services and ETL pipelines over high-volume logistics data, with tested REST APIs for real-time analytics.
- Cut delivery lead time by 30% at peak load by tuning Redis caching and distributed task scheduling to improve service latency and throughput.
- Worked through code reviews, design discussions, and planning sessions to make the data and service boundaries easier to operate as the system grew.
- Python
- ETL
- REST APIs
- Redis
Things I've built, and why.
All projects01 / VITAL
Request → Route → Plan → Approve → Commit
AI systems / Control
An AI planner that asks before it acts.
A planning assistant with checkpointed state, specialist routing, and human approval before a calendar write.
The key decision
Make approval part of the execution graph.
Python · LangGraph · FastAPI · Postgres
Read case study →02 / Research Paper RAG
Upload → S3 → SQS → Process → Retrieve
Cloud infrastructure / Retrieval
From an uploaded paper to a checkable answer.
An AWS research assistant with asynchronous document processing, per-user retrieval, and answers linked to sources.
The key decision
Move expensive processing behind a queue.
AWS Lambda · SQS · Bedrock · Pinecone
Read case study →03 / SmartCache AI
Ingest → Content pool → Recommend → Queue → Download
Local AI / Background systems
Useful content, ready before the signal drops.
A commute-aware content pipeline combining deterministic ingestion, local-model recommendations, and background downloads.
The key decision
Use agents only where judgment is needed.
Django · AutoGen · Ollama · Celery
Read case study →
The tools behind the work.
From projects and experience. Select any skill to see where I used it.
Languages
Python
Backend services, agent workflows, ingestion handlers, and machine-learning experiments across my projects.
Backend & web
Cloud & infrastructure
AI & machine learning
Data & storage
Deployment & developer tools
Systems & engineering practices
Product & collaboration
Work that can be inspected outside this portfolio.
Two co-authored papers trace one research question from literature review to implementation. Two upstream contributions show how I work inside codebases I did not start.
Research
Model repository ↗IJRASET · Volume 11, Issue XI
2023
Reviewed unimodal approaches across facial, textual, audio, EEG, and physiological signals, then proposed a combined facial, text, and HRV direction.
Read the paper ↗Bulletin for Technology and History Journal · Volume 24, Issue 5
2024
Carried the research direction into an implemented framework using facial, questionnaire/text, and HRV components; the paper reports 82% accuracy for its happy/sad image classifier and discusses the remaining false-positive and false-negative trade-offs.
Read the paper ↗
Merged upstream
Responsible Alliance Protocol
Hardened Kubernetes manifests for an OPA policy service
Added namespace isolation, quotas and limits, hardened replicas, read-only policy configuration, and default-deny network policies with explicit access rules.
View merged PR ↗Developer Solutions Lab
Offline AWS region detection for on-prem readiness
Built a TypeScript diagnostic that resolves local AWS region configuration without network access or credentials.
View merged PR ↗
Let’s talk about the work.
For software engineering opportunities in backend, cloud infrastructure, and AI systems.