Syed Basit Abbas

Syed Basit Abbas

Founding Engineer specializing in Full-Stack AI, ML Pipelines, and Multi-Agent Systems

Based in: Lahore, PakistanCurrently: Senior AI Engineer, Mod SolutionsExperience: 3 years

Hire me

Syed Basit Abbas is a versatile founding engineer with extensive experience building end-to-end AI products using React, Python, and advanced ML frameworks like PyTorch and TensorFlow. He has a proven track record of deploying production-ready AI agents, RAG systems, and computer vision models for startups across the US, Canada, and the UAE.

Experience

Senior AI Engineer

Mod SolutionsUAE (Remote)workApr 2026 – Present

Summary

Built and deployed an AI property-concierge system (Next.js, TypeScript, OpenRouter) with streaming responses and LLM tool-calling under a manager-agent architecture, running in production on web and WhatsApp. Built retrieval on MongoDB Atlas using embeddings and vector search, so the assistant answers only from live property inventory instead of made-up listings, and tested it against real queries to improve answer accuracy. Led the AI architecture, selecting models, retrieval approach and agent structure, and analysing output quality to find where accuracy, latency and reliability could improve. Integrated third-party APIs across the platform, including Meta Ads Conversions API tracking and WhatsApp messaging, and shipped the back end on Django REST.

Results

  • Grounded assistant answers in retrieval over MongoDB Atlas with embeddings and vector search to ensure the assistant only cited live inventory instead of hallucinating listings.
  • Reduced perceived response time by implementing streaming responses for web and WhatsApp interfaces.
  • Identified and resolved system failure cases by testing the architecture against real user queries.
Next.jsTypeScriptOpenRouterLLM tool-callingMongoDB AtlasEmbeddingsVector searchMeta Ads Conversions APIWhatsApp messagingDjango REST

Co-Founder

PrepzyCanada (Remote)founderFeb 2026 – Present

Summary

Co-founded Prepzy, a platform for AP exam prep (Calculus MCQ and free-response question banks), and led its AI and back-end architecture. Developed and deployed the back end on FastAPI and MongoDB Atlas (async Motor driver) with cookie-based auth and bcrypt hashing, running serverless in production. Built the React front end (Radix UI, TanStack Query, Google OAuth, React Hook Form) wired to the FastAPI back end.

What I did

  • Focused on question completion rates and core retention

Results

  • Reached roughly 300 active students preparing for AP Calculus
FastAPIMongoDB AtlasMotor driverCookie-based authbcryptReactRadix UITanStack QueryGoogle OAuthReact Hook Form

Founding Engineer

RolyncUnited States (Remote)founderJul 2025 – Apr 2026

Summary

Built Rolync, an AI referral-matching system on LangGraph and MongoDB using a multi-agent architecture to match candidates to referrals. Ran the agents on Gemini and Perplexity with prompt engineering tuned for career-goal alignment, and used RAG over MongoDB for scalable retrieval.

What I did

  • Validated relevance and matching accuracy for the early-stage build.

Results

  • Built a multi-agent matching engine that facilitated around 1k+ referral introductions.
LangGraphMongoDBGeminiPerplexityPrompt engineeringRAG

Skills

Technical

Gemini
PythonPython
LLM integration
RAG
AI agents
Prompt engineering
OpenAI
TypeScriptTypeScript
Data preprocessing
Embeddings
Next.js 15Next.js 15
Tool calling
Tailwind CSSTailwind CSS
Model training and fine-tuning
Computer vision
GitHubGitHub
Anthropic
FastAPIFastAPI
Supervised and unsupervised learning
JavaScriptJavaScript
Cursor
PyTorchPyTorch
GitGit
REST APIs
MongoDB AtlasMongoDB Atlas
Pandas
NumPy
React 19React 19
Serverless deployment
SQLSQL
Node.jsNode.js
Django RESTDjango REST
Authentication
Evals
Feature engineering
Walk-forward validation
CNNs
Transformers
OpenCVOpenCV
NLP
TensorFlowTensorFlow
Keras
scikit-learn
Voice agents
RetellAI
Ingestion pipelines
MySQLMySQL
VercelVercel
Jupyter
Claude Code
DockerDocker
AWS
GCPGCP
CI/CD
DeepFace
RNNs and LSTMs
Time-series modeling
Kaggle
Function calling
AOE4World API
LangGraph
Gaussian HMM

Projects

AOE Master

Developerpersonal
Real-time AI coach for Age of Empires IV using the AOE4World API and Gemini function-calling.
AOE4World APIGemini function-calling

AI Trading Bot

Developerpersonal
Machine learning system in Python with feature engineering on market data, an unsupervised Gaussian HMM for regime detection, a backtester using walk-forward validation to measure out-of-sample performance, and a React monitoring dashboard.
PythonGaussian HMMReact

Face Recognition Validation

Developerpersonal
Computer vision identity-verification service in Python using DeepFace image embeddings, evaluated across varied lighting conditions.
PythonDeepFaceImage embeddings

AI Voice Agent

Developerpersonal
Voice AI calling agent on RetellAI, wired through Make.com to a database, that identifies new or returning callers and handles their queries in real time.
RetellAIMake.com

Mitrat

Developerpersonal
NLP and RAG system for an NDIS provider (Python, LangChain) combining SQL data with a document ingestion pipeline behind a chat interface.
PythonLangChainSQL

Wrapped by Keeya

Developerpersonal
Full-stack e-commerce store (Next.js 15, React 19) with Stripe and Klarna checkout.
Next.js 15React 19StripeKlarna
Visit the site

Custom Walk-Forward Backtester

personal
Built a custom backtester from scratch in Python to test sample performance without data leakage. Implemented strict rolling window slices manually to avoid future data leaks common in standard libraries.
PythonGaussian HMM

MeddAI

Developerpersonal
Multi-agent service on LangGraph that evaluates each model's output across OpenAI, Gemini and Perplexity and routes tasks to the most accurate one.
LangGraphOpenAIGeminiPerplexity

Education

BS Artificial Intelligence

University of Management & Technology (UMT)Artificial Intelligence

Societies & activities

Dean's Merit Award

Recognition

Awards

Dean's Merit Award

University of Management & Technology (UMT)
Hire SyedReach out about a role, a contract or a conversation.For recruiters

Syed's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis