The model is the easy part
Most of the difficulty in an AI system sits upstream and downstream of inference — data correctness, retrieval quality, latency budgets, failure modes. That is where the work actually is.
Open to AI Engineer and Software Engineer roles
I build AI systems that hold up in production.
4+ years across research, healthcare, and financial infrastructure — most of it on the parts of a system that decide whether a model is actually useful.
- Building AtriveoJob-search platform · 100+ users
- MS Data ScienceStony Brook · May 2026
- Financial NLP researchStony Brook University
- Open to new rolesAI/ML · SWE · Data
The short version
Atishay Kasliwal is an AI Engineer based in New York. He builds retrieval-augmented generation systems, LLM agents, and the event-driven infrastructure they run on — the unglamorous part where latency budgets, cost per query, and failure modes decide whether a model is actually useful.
He holds a Master of Science in Data Science from Stony Brook University and a Bachelor of Technology in Computer Science and Information Technology from Symbiosis University of Applied Sciences. Before graduate school he spent three years at Accolite Digital as a Senior Software Engineer, shipping event-driven ETL and serverless platforms for Fidelity Investments and BT Group.
He currently builds and operates Atriveo, a job-search platform serving live customers, and researches financial NLP at Stony Brook University.
The long version is on the experience page, role by role, with the numbers attached. Peer-reviewed work is listed under research.
Work has run at
- Fidelity Investments
- BT Group
- Stony Brook University
- Wake Forest University
- Accolite Digital
Focus
What I work on
- LLM Systems
- RAG
- AI Agents
- Distributed Systems
- Developer Tools
- Cloud Infrastructure
- Applied Research
Method
How I work
Measure the thing you claim
A number without a definition is decoration. P99 under what load, accuracy on which split, cost per query at what volume. If I cannot say how it was measured, it does not go on the slide.
Build for the failure case
Production systems are defined by what they do when a dependency is slow, a schema changes, or a model returns nonsense. The happy path is the demo; the rest is the job.
Ship, then earn the complexity
Every abstraction should be paid for by a problem that already exists. I would rather ship something plain and add structure under real pressure than build for a scale that never arrives.
Selected work
Things I have shipped
- Product
Atriveo
Job-search platform with a 5.0-rated Chrome extension and live customers.
- 100+
- active customers
- 2K+
- daily queries
- Applied Research
FOMC Intelligence
NLP pipeline turning Federal Reserve communications into market signals.
- 200K+
- data points processed
- 3.7K+
- time-series intervals
- AI Systems
Legal RAG
Retrieval-augmented generation over legal filings, with citations that hold up.
- Citation-first
- answer contract
- Chunk-level
- source attribution
Depth
Technical range
AI & Machine Learning
- Large Language Models
- Retrieval-Augmented Generation
- LLM Agents
- PyTorch
- Transformers
- LangChain
- Vector Search
- Embeddings
- NLP
- Computer Vision
- scikit-learn
- Model Evaluation
Languages
- Python
- TypeScript
- JavaScript
- Java
- SQL
- Go
- C++
Backend & Distributed Systems
- FastAPI
- Spring Boot
- Node.js
- Kafka
- REST
- GraphQL
- Microservices
- Event-Driven Architecture
- OAuth 2.0
- Caching
- Message Queues
Data & Storage
- PostgreSQL
- DynamoDB
- MongoDB
- Redis
- Elasticsearch
- MySQL
- pgvector
- ETL / ELT
Cloud & Infrastructure
- AWS (Lambda, ECS, S3, SQS, API Gateway)
- GCP
- Cloudflare Workers
- Docker
- Kubernetes
- Terraform
- Jenkins
- CI/CD
- Prometheus
- CloudWatch
Frontend
- React
- Next.js
- TypeScript
- Vite
- CSS Architecture
- Web Performance
- Accessibility
Education
Where I studied
Master of Science in Data Science
Stony Brook University · Stony Brook, New York
Neural Networks · Big Data Algorithms and Networks · Big Data Analytics · Statistical Computing · Data Management
Bachelor of Technology in Computer Science and Information Technology
Symbiosis University of Applied Sciences · Indore, Madhya Pradesh, India
Distributed Systems · Operating Systems · Computer Networks · Data Structures · Algorithms · Software Development
Elsewhere
Outside the terminal
Photography
Street and travel work, shot on a Nikon.
Writing
Notes on what breaks in production ML.
Open source
Public repositories, pulled live from GitHub.
FAQ
Common questions
Who is Atishay Kasliwal?
Atishay Kasliwal is an AI Engineer based in New York who builds production large language model systems, retrieval-augmented generation pipelines, and the distributed infrastructure they run on. He holds a Master of Science in Data Science from Stony Brook University.
What does Atishay Kasliwal specialize in?
Large language model systems, retrieval-augmented generation, AI agents, event-driven distributed architecture, and cloud infrastructure on AWS and GCP. His primary languages are Python, TypeScript, and Java.
Where did Atishay Kasliwal study?
Atishay Kasliwal earned a Master of Science in Data Science from Stony Brook University and a Bachelor of Technology in Computer Science and Information Technology from Symbiosis University of Applied Sciences in Indore, India.
Where has Atishay Kasliwal worked?
He is a Graduate Research Assistant at Stony Brook University, previously a machine learning intern at the Wake Forest University Center for Artificial Intelligence Research, and spent three years as a Senior Software Engineer at Accolite Digital building systems for Fidelity Investments and BT Group.
Is Atishay Kasliwal available for hire?
Open to AI Engineer and Software Engineer roles. Available immediately · Open to relocation · Authorized to work in the US. He can be reached at [email protected].
Open to work
Currently at Stony Brook University, and open to what is next.
Available immediately · Open to relocation · Authorized to work in the US