Open to Full-Time AI Engineering Roles — May 2026

TejaswiniChavan

AI Full-Stack Engineer & Cognitive Systems Researcher

"Architecting production-grade AI systems — directed-graph multi-agent frameworks, knowledge graphs at 95% precision, and memory threading across 535+ moments. I close the gap between research and reality."

⚡ Hire Me ⬇ Download Resume

Years Engineering

Queries/sec Production

Knowledge Graph Precision

AI Systems Shipped

Core AI/ML Technology Stack

Python

TypeScript

React

AWS

Terraform

Docker

PostgreSQL

MongoDB 🧠 Multi-Agent · RAG · LLMs

Python

React

AWS

TypeScript

Docker

Who I Am

Engineer. Researcher. AI Architect.

Years Engineering

95%

Knowledge Graph Precision

200+

Queries/sec

AI Systems Shipped

I am a high-impact AI Full-Stack Engineer with 4+ years of engineering experience shipping production-grade multi-agent systems, knowledge graphs, and cloud-native AI infrastructure at the intersection of cognitive computing and enterprise software.

At NGI, Inc, I lead the ML Agent Integration Layer as a Graduate Research Assistant where I'm architecting a directed-graph multi-agent framework with a shared state bus, cross-session memory threading via PostgreSQL BFS traversal + Hugging Face Sentence-Transformers (86% thread connectivity across 535 moments), and a production knowledge graph achieving 95% precision at 200+ queries/second.

My engineering foundation spans 2.5 years at Avalara building Java/Spring Boot APIs, Salesforce LWC platforms, Oracle Eloqua integrations, Apache Kafka pipelines, and enterprise-scale CI/CD at production scale — combining software engineering precision with AI research rigor from my MS at Northeastern University.

I hold AWS Solutions Architect Associate and Salesforce AI Specialist certifications, and have published technical research on Memory Threading for the Humanitarians AI ecosystem.

Multi-Agent SystemsRAGOpenAI APIVector DBsAWS · TerraformFastAPIspaCy · BERTSalesforce LWCApexSupabase

AI / ML

PyTorch

TensorFlow OpenAI API LangChain RAG pgvector spaCy BERT GPT-4o

Backend

Python FastAPI

Java

Spring Boot

Node.js Express

TypeScript

Cloud & DevOps

AWS

Terraform

Docker

Kubernetes App Runner Lambda CloudFront

Data & Storage

PostgreSQL

MongoDB

Redis

Firebase Supabase Pinecone

Salesforce Ecosystem

Salesforce DX LWC Apex Visualforce SOQL · SOSL REST APIs Custom Objects AI Specialist Certified Platform Developer

Career

Professional Experience

Graduate Research Assistant

Northeastern University × NGI, Inc · Boston, MA

Jan 2026 — Apr 2026

Engineered a directed-graph multi-agent architecture with three specialized agents (Priority, Theme Organizer, Loose End Detector) sharing a common state bus — achieving 500ms end-to-end latency on the NLP ingestion pipeline.
Architected OpenClaw autonomous agent framework integration for real-world task execution inferred from conversational context, enabling proactive delivery without explicit user prompts.
Refactored and integrated SynthesisAgent into a CrewAI researcher agent, improving task completion through cross-session memory context and plain-language thread insights.
Full production infrastructure on AWS with Terraform — App Runner, RDS, Lambda, S3, CloudFront — designed for zero-downtime deployment.

PythonFastAPICrewAIOpenAISupabaseAWSTerraformMulti-Agent

AI/ML Engineer Intern (Co-op)

NGI, Inc (Stealth Startup) · Boston, MA

Sept 2025 — Dec 2025

Orchestrated a production-scale knowledge graph using fuzzy matching, vector similarity search, and rule-based classification — achieving 95% precision at 200+ queries/second.
Spearheaded complete Memory Threading system using PostgreSQL graph schema + BFS traversal on Hugging Face Sentence-Transformers — 86% thread connectivity across 535 moments with 1,024 relationships.
Designed hybrid NER pipeline integrating spaCy and BERT with multi-signal scoring, linking ambient voice moments to calendar and email metadata — 78% keyword coverage across production Supabase infrastructure.

PythonspaCyBERTHuggingFacePostgreSQLKnowledge GraphsSupabase

AI Software Engineering Fellow

Headstarter AI · Remote

Jul 2024 — Aug 2024

Built AI customer support chatbot using LangChain, Pinecone, and AWS Bedrock with RAG pipeline and real-time knowledge base retrieval — improved resolution accuracy 35%, deployed at 99.9% uptime.
Developed RAG-powered professor recommendation system (Next.js, OpenAI API, Pinecone) with Python scraping + vector ingestion pipeline; semantic search across 90+ user reviews.

LangChainPineconeAWS BedrockOpenAIRAGNext.js

Software Engineer

Avalara Technologies · Pune, India

Jul 2021 — Dec 2023

Built 3 Salesforce Lightning Web Component (LWC) interfaces for Sales and Marketing teams — 35% decrease in data retrieval time and boosted report creation speed; developed Apex classes, triggers, and custom objects for enterprise data integration.
Integrated Oracle Eloqua marketing automation with Salesforce CRM — automated lead scoring, nurturing sequences, and campaign attribution workflows, enabling real-time sync of marketing and sales pipeline data.
Created intelligent Node.js data pipelines with ML-based anomaly detection and Redis caching — reduced API response times by 200ms and improved Web Help functionality through JavaScript data validation.
Implemented Java/Spring Boot applications with Postgres — 35% increase in data processing speed; automated tax compliance workflows with rate limiters and RESTful APIs — 70% reduction in processing time.
Resolved critical CI/CD pipeline issues using Packer, GitHub, and Amazon CodeDeploy — 30% reduction in deployment failures and improved overall team velocity.

JavaSpring BootNode.jsSalesforce LWCApexOracle EloquaSOQLMongoDBRedisAWSCI/CD

Software Engineering Intern

Avalara Technologies · Pune, India

Dec 2020 — Jun 2021

Constructed Java/Spring Boot application with Postgres relational database — 35% increase in data processing speed for key reporting features.
Built Python + Apache Kafka data pipelines for real-time analytics with SQL data validation — 99.9% data accuracy across dashboards and interactive visualizations.
Developed A/B testing framework with UX design team — improved user engagement by 20% and enhanced conversion rates on key product features.
Designed Sentry error logging system providing real-time alerts and diagnostics — team resolved critical issues 60% faster than before.

JavaSpring BootPythonApache KafkaSQLPostgresSentryA/B Testing

Selected Work

Key Projects

01 / PROJECT

Financial Advisor Copilot

⚡ 6 hrs/week saved per advisor

AI copilot synthesizing complex financial data into actionable insights — a three-agent pipeline (Access, Connection, Summary) integrating live Salesforce CRM via SOQL. Human-in-the-loop accountability layer generating unified AI briefings in under 60 seconds.

Salesforce CRM→3-Agent Pipeline→Human-in-Loop→AI Briefing <60s

PythonFastAPISalesforce SOQLReact/TypeScriptDockerStreamlit

🔗 GitHub

02 / PROJECT

ML Agent Integration Layer

⚡ Production · Node Startup

Self-evolving multi-agent memory system. Three FastAPI agents + ProactiveDelivery + SynthesisAgent (gpt-4o-mini). Vector retrieval, per-user feedback weights, cross-session persistence.

Moments→NLP Pipeline→Priority Agent→Supabase

PythonFastAPIOpenAIpgvector

🔗 GitHub

03 / PROJECT

SmartScreen ATS

⚡ Semantic Matching

OpenAI embeddings + GPT-3.5 over 1536-dim vectors. Top 5–10 candidates auto-ranked by cosine similarity — smart hiring, not keyword bingo.

Resume→OpenAI Embed→Cosine Rank→Top Matches

OpenAIRAGPython

🔗 GitHub

04 / PROJECT

AutoQA Multi-Agent System

⚡ LLM-Powered QA

Planner, Executor, Verifier, Supervisor agents collaborating like a real QA team. Agents-S + LangChain, Android simulation.

Test Case→Planner→Executor→Verifier

LangChainMulti-AgentPython

🔗 GitHub

05 / PROJECT

CloudNative IaC

⚡ 20% Faster Deploy

Serverless AWS via Terraform + Packer AMI automation. App Runner, Lambda, RDS, S3, CloudFront — all infra reproducible as code.

Code Push→Terraform→Packer AMI→AWS Deploy

AWSTerraformPacker

🔗 GitHub

06 / PROJECT

Pantry Tracker AI App

⚡ 92% Image Accuracy

GCP Vertex AI AutoML classifier. Next.js + Firebase. Groq LLaMA for real-time recipe suggestions from expiring ingredients.

Photo→Vertex AI→Item ID→LLaMA Recipe

Next.jsGCP Vertex AILLaMA

🔗 GitHub

07 / PROJECT

AI Customer Support Chatbot

⚡ RAG + Claude + GPT-4

RAG-enhanced chatbot with LangChain + Pinecone, AWS Bedrock (Anthropic Claude, GPT-4). Multi-model, context-aware support.

Query→Pinecone RAG→Claude/GPT-4→Response

RAGLangChainPinecone

🔗 GitHub

Thought Leadership

Technical Insights & Industry Thinking

Follow on Medium Articles on AI systems, agent design & cloud architecture

📝 Article · Medium

Shopping Cart Abandonment: Challenges & Solutions for CVS Online Drugstore

UX research deep-dive into why customers abandon carts on CVS's online pharmacy — identifying friction points and proposing redesign solutions: guest checkout, pricing transparency, and optimised "Add to Cart" flows.

8 min read→ Read on Medium

💡 Guide · LinkedIn

Building Self-Evolving Agent Systems: Lessons from V1 to V3

How per-user signal weights, vector retrieval, and SharedContext turned a simple FastAPI service into an adaptive multi-agent system.

9 min read→ Read on LinkedIn

🎯 Case Study · LinkedIn

AWS Infrastructure Consolidation with Terraform: A Real Startup Story

How I designed and pitched a full AWS cloud consolidation for a stealth wearable startup — Terraform IaC from zero to production.

6 min read→ Read on LinkedIn

🔬 Opinion · LinkedIn

Why Your RAG System Needs Per-User Signal Weights

Generic RAG retrieves the same results for everyone. Personalizing signal weights through feedback loops is the real unlock for intelligent AI.

5 min read→ Read on LinkedIn

Thought Leadership

LinkedIn Technical Insights & Posts

#MultiAgent

How I Evolved a Single FastAPI Agent Into a Self-Coordinating System

The SharedContext state bus design that made cross-agent coordination seamless — and what broke along the way from V1 to V3.

→ Read on LinkedIn

#RAG

Vector Retrieval vs Brute-Force: A Real Production Comparison

Why replacing brute-force scanning with pgvector retrieval changed what kinds of queries were even possible in our agent service.

→ Read on LinkedIn

#AWS · #IaC

Terraform Lessons From Consolidating a Startup's Cloud in 4 Weeks

App Runner + Lambda + RDS + CloudFront from scratch — the state management pitfalls nobody warns you about.

→ Read on LinkedIn

#MemoryAI

The Future of AI Is Memory — Here's Why I'm Building It

Reflections on building a system that doesn't just answer questions but remembers context, surfaces patterns, and gets smarter over time.

→ Read on LinkedIn

My Approach

Design Philosophy & Architecture Thinking

🧠

Memory-First AI

AI without memory is expensive autocomplete. I build agents that accumulate context, surface patterns, and grow smarter — across sessions, not just within them.

✓ Memory Threading · Cross-session persistence · Self-evolving weights

⚡

Ship Fast, Break Nothing

Production is unforgiving. Every decision prioritizes API contract stability, graceful fallbacks, zero-downtime deploys. When a pipeline depends on my service, it never goes down.

✓ API contract stability · V1→V2→V3 zero regressions

🔬

Research → Reality

I close the gap between research and production. Ablation studies, formal evaluation — applied directly against real user data in a real startup, not a sandbox notebook.

✓ MS Research + Production co-op simultaneously

☁️

Infrastructure as a Superpower

Most AI engineers ignore infra until it breaks. I treat Terraform, AWS, and CI/CD as first-class skills that multiply every AI system I build.

✓ AWS Activate · Terraform IaC · App Runner + Lambda + RDS

Growth Milestones

Career & Impact Timeline

2021

First Production APIs @ Avalara

Kafka pipelines (99.9% accuracy), Spring Boot APIs, Salesforce LWC. 70% reduction in tax processing time.

Full-Time Engineer

2023

MS @ Northeastern — Pivoted to AI

Left Avalara. MS Information Systems. Full focus on AI engineering and LLM systems.

Academic + Research

2024

Built Foundational Recall Engine

Lifestyle classifier 100% accuracy, 213+ moments/sec. Hybrid NER (spaCy + BERT). Fleet Simulation Suite.

Co-op @ Node

2025

Shipped ML Agent Layer V1 → V3

Multi-agent system: vector retrieval, self-evolving weights, ProactiveDeliveryAgent, SynthesisAgent. All in production.

3 Versions Live

2025

AWS Infrastructure Consolidation

Terraform IaC — App Runner, RDS, Lambda, S3, CloudFront. Won AWS Activate Founders program.

AWS Certified

🏗

May 2026

Graduating — Seeking Full-Time

MS Information Systems, Northeastern. Targeting AI Engineering at companies building real AI systems.

Available Now

TejaswiniChavan

Engineer. Researcher. AI Architect.

Technical Skills

Core AI/ML

Engineering

Cloud & DevOps

Professional Experience

Key Projects

GitHub Activity & Contributions

Technical Insights & Industry Thinking

Shopping Cart Abandonment: Challenges & Solutions for CVS Online Drugstore

Building Self-Evolving Agent Systems: Lessons from V1 to V3

AWS Infrastructure Consolidation with Terraform: A Real Startup Story

Why Your RAG System Needs Per-User Signal Weights

LinkedIn Technical Insights & Posts

How I Evolved a Single FastAPI Agent Into a Self-Coordinating System

Vector Retrieval vs Brute-Force: A Real Production Comparison

Terraform Lessons From Consolidating a Startup's Cloud in 4 Weeks

The Future of AI Is Memory — Here's Why I'm Building It

Design Philosophy & Architecture Thinking

AI System Capabilities & Expertise

Career & Impact Timeline

Certifications & Achievements

Education & Training

My Resume

Let's build something that matters.