AI Engineer — New York City

I turn AI prototypes into systems that hold up in production.

Currently building agentic verification and recruiting pipelines at Centh AI — parallel LLM agents processing 800+ documents and 500+ emails a day. Previously cut an AI assistant's response latency by more than half at Novum AI.

View the work Résumé ↓

Suggestion Latency

3.4s → 1.5s

Docs Verified / Day

800+

Emails Automated / Day

500+

Prompt Tokens Cut

−60%

Experience — 01 / 03

AI Engineer

Centh AI (South Park Commons-backed) · May, 2026 – Present

Forward-deployed across enterprise clients on an agentic document-credentialing pipeline — a database-backed queue orchestrates parallel verification agents, turning a manual review bottleneck into a same-day recruiter handoff.

Parallel verification agents

Claude Haiku (LLM checks) and Mistral (OCR) run concurrently on every document

Forward-deployed ownership

End-to-end system design across enterprise clients, translating live requirements into production decisions

Candidate engagement agent

Coordinates clinical test scheduling, like TB screenings, ahead of onboarding

Recruiting + consolidation

500+ emails/day screening agent with a human-feedback loop; merging two production codebases into one architecture

Centh AI's model: don't rip out the legacy system a company has years of data in — put an agent directly on top of it. The dashboard is just configuration; the agents do the real work.

Claude APIMistralPythonGCPNext.jsTypeScriptPostgreSQLAgentic Systems

Documents Verified / Day

800+

2 parallel agents

Verification pipeline — Haiku + Mistral, parallel

Experience — 02 / 03

AI Software Engineer

Novum AI · Jun, 2025 – Apr, 2026

Owned the retrieval and generation path for an AI call assistant surfacing live suggestions mid-conversation — every millisecond of lag was a rep waiting on the model.

Parallelized retrieval

Concurrent multi-source lookups instead of sequential fetches

Hierarchical chunking

Layout-aware segmentation that preserved document structure

Hybrid search

Dense embeddings combined with BM25 keyword matching

Prompt compression

Context trimmed from 4,900 to 1,950 tokens per call

AWS LambdaDynamoDBAWS SAMAssemblyAIAWS BedrockWebSocketsFastAPIJWTRBACRedisCI/CDServerless Architecture

Suggestion Latency

3.4s

1.5s

−56% faster

Retrieval pipeline — parallelized

Experience — 03 / 03

Software Engineer

InquisAI (Startup Project) · Jun, 2024 – Mar, 2025

Built an AI assistant platform end-to-end as one of three engineers, from semantic retrieval to the async backend serving it.

Developed an AI assistant platform using LangChain, OpenAI Embeddings, and Chroma vector store, achieving accurate document retrieval on 500+ internal test queries with semantic search

Redesigned backend by migrating Flask to FastAPI with asynchronous processing and optimized database queries, reducing API latency by 30% through performance testing

Led Agile development in a 3-person team, using Azure DevOps to manage sprints and implementing CI/CD pipelines for automated testing and deployment to AWS

FastAPILangChainOpenAIChromaVector DatabaseSemantic SearchAzure DevOpsFlaskPostgreSQLDockerRESTful APIsAsync ProcessingCI/CD

30% latency reduction

Education

Grounded in the fundamentals.

New York University

2023 – 2025

Master of Science in Computer Science

New York, NY, USA

3.8/ 4.0 GPA

University of Mumbai

2019 – 2023

Bachelor of Engineering in Computer Engineering

Mumbai, Maharashtra, India

9.0/ 10.0 GPA

Personal Projects

What I build on my own clock.

Independent systems built end to end — from an AI-native commerce platform to the infrastructure fixes most tutorials skip.

In Progress

Stoca

AI-Native Local Commerce Platform

Built an AI-native commerce platform where store owners manage their business through conversation with a Claude-powered assistant supporting 19 tool calls via Vercel AI SDK streaming, covering inventory, pricing, orders, and promotions. Integrated Claude Vision, pgvector semantic search, and the Pexels API into an autonomous product-enrichment pipeline for shelf-photo identification and catalogue matching.

19tool calls, one conversation

Shelf-photo enrichment pipeline

Autonomous enrichment — photo to catalog, no manual entry

Next.jsTypeScriptFastAPIPythonPostgreSQLSupabasepgvectorClaude APIGitHub Actions

Complete

FlashBids

Anti-Sniping Auction Platform

Created a real-time auction platform using Flask and WebSockets with automatic time extensions to prevent last-second bid sniping, deployed on AWS EC2 with Auto Scaling and DynamoDB. Profiled a Redis connection bottleneck (50-80ms per new connection during bid floods) and implemented connection pooling, dropping p95 latency from 800ms to 200ms.

P95 Bid Latency

800ms

200ms

−75% faster

AWS EC2DynamoDBS3RedisWebSocketsCloudWatchFlaskPythonREST APIsDevOps

Live Demo ↗Source ↗

Also Shipped

Inquis AI

Spearheaded development of an Al assistant builder leveraging LangChain and OpenAI Embeddings to vectorize documents for domain-specific responses via GPT-40. Redesigned backend from Flask to FastAPI, cutting API latency by 30%. Deployed scalable RESTful APIs supporting 1K+ concurrent users and led Agile sprints.

FastAPI · Python · LangChain · OpenAI

Demo ↗

Home Store

Designed a microservices architecture using Spring Boot, Eureka service discovery, and API Gateway, with modular services enabling independent deployment and horizontal scaling across tenants. Containerized services with Docker and Docker Compose and exposed REST APIs with SwaggerUI documentation, enabling consistent testing and rapid iteration.

Spring Boot · Spring Cloud · Eureka · Docker

Source ↗

Smart Photo Album

Created a web application for uploading, storing, and searching photos using natural language queries. Leveraged AWS ElasticSearch, S3, and Lambda for automation and scalability.

AWS S3 · ElasticSearch · Lambda · Lex

Demo ↗Source ↗

VibeCheck

Created Django/Python social platform pairing users via Spotify data. Established reactive messaging with Redis Pub/Sub (30-40% latency reduction). Deployed on AWS Elastic Beanstalk with CI/CD (87% test coverage).

Python · Django · Redis · PostgreSQL

Demo ↗Source ↗

Roomio

Built an apartment rental and roommate finder web application with a secure and scalable architecture. Designed an optimized relational database schema for efficient data transactions.

Django · Python · PostgreSQL · HTML

Source ↗

Leadership

Invested in other people.

Teaching, building, and organizing outside the day job.

Software Teaching Assistant

New York University · New York, NY · Jan, 2024 – May, 2025

Mentored 50+ graduate and undergraduate students in C++ and Java, focusing on object-oriented design, data structures, and clean coding practices

Reviewed student code and provided detailed feedback on correctness, design choices, and performance trade-offs

Assisted students with design patterns, SOLID principles, and translating theory into working implementations

Provided career and technical guidance beyond coursework, helping students understand industry expectations and technology choices

C++JavaOOPDesign PatternsSOLID PrinciplesCode ReviewsMentorship

Students Mentored

50+

Multiple semesters

Student Web Developer

Google Developer Student Club BVCOE · Navi Mumbai, India · Aug, 2021 – July, 2022

Developed and launched a website for the Google Developer Club with a team of 5 members using the React.js framework, resulting in a 20-30% increase in club visibility and member engagement

Organized key events in cloud computing, web, and Android development, resulting in a virtual workshop with over 300 attendees

Showcased interactive tech learning experiences and fostered community engagement through technical workshops

ReactJavaScriptWeb DevelopmentEvent Management

Workshop Attendees

300+

+20-30% engagement

Certifications

Coursework worth naming.

Technical Interview Prep

View ↗

Data Structures · Algorithms · Problem Solving

CodePath · 2024

Introduction to Data Science in Python

View ↗

Python · Numpy · Pandas · Data Cleansing

University of Michigan (Coursera) · 2023

Version Control

View ↗

GitHub · Bash (Unix Shell) · Web Development · Linux

Meta (Coursera) · 2023

Principles of UX/UI Design

View ↗

Figma · Web Design · Accessibility · Front-End Web Development · User Experience (UX)

Meta (Coursera) · 2023

Stack

The stack behind the systems.

Sized by how load-bearing each tool is in production, not by how many logos fit on a page.

Core to productionUsed in productionFamiliar

01Languages: PythonTypeScriptJavaC++
02Frontend: ReactNext.jsTailwind CSSJavaScript
03Backend: FastAPILangChainFlaskSpring BootDjangoNode.jsRESTful APIs
04Cloud & DevOps: AWSDockerAWS LambdaAWS SAMGCPDocker ComposeGitHub ActionsSwagger / OpenAPIKubernetesTravis CI
05Databases: PostgreSQLDynamoDBSQLRedisSupabaseMySQLMongoDBPineconeFirebase
06System Design: Agentic SystemsMicroservicesEvent-Driven ArchitectureDistributed Systems
07Testing & Monitoring: PyTestCloudWatchJUnitDjango TestPostHogChrome DevTools
08Tools & Workflow: GitGitHubFigmaAgile (Scrum)

Domains Covered

Tools Catalogued

48+

Core to Production

Python · FastAPI · AWS · Next.js · Agentic Systems

Get In Touch

Let’s build something that holds up.

Open to AI engineering roles and collaborations. Reach out anytime.

Open to opportunities · New York City