Babar Hashmi
Book a call (opens in new tab)
Portrait of Babar Hashmi

Babar Hashmi

AI Engineer - Agents, LLMs, Computer Vision

Available remote - globally

Bangalore, IndiaOpen to relocation

AI engineer with 3 years of applied ML, most of it as the first AI hire at QuillAudits (now QuillAI Network). I started with no team and no existing pipeline, and built the models, agent workflows, and developer SDK that the company now ships. Day to day that means fine-tuning LLMs for detection tasks, designing agent loops, and writing the docs and tests that keep results honest.

Book a callDownload CV

Now

Founding AI engineer at QuillAI Network, building agent systems from first prototype to production.

Impact

+25%
higher detection precision on labeled contracts
60%
less manual review time on audits
93%
accuracy on ERC20 safety checks
3
business units shipping on the developer SDK
100K+
smart contracts analyzed to date

Measured on the QuillAI audit pipeline: detection-model precision on labeled contracts, review hours before and after rollout, 100K+ contracts scanned to date.

GitHub

GitHub avatar of ibabarhashmi@ibabarhashmi
DesiRAG
Hindi voice RAG over multilingual search passages: hybrid retrieval, guardrails, sub-20ms latency
Python
organon
Is this DeFi yield real, and what could kill it? An ML scorecard, not an opinion.
TypeScript
GoldenHour
First response checklist for cyber fraud
TypeScript
Agentic-Auditor
On-the-go AI smart contract auditor
TypeScript
Showing cached data - live stats load in the browser.
View all on GitHub ↗

Experience

Founding AI Engineer (MTS, Research Engineering)

Jul 2024 - Present
QuillAI Network (QuillAudits)
  • Shipped production deep-learning models for smart-contract vulnerability detection, hardened by red teaming. They catch what static analysis misses.
  • First AI hire: set up the research-to-production process, then designed agent workflows with MCP tool calling on open agent frameworks and shipped a developer SDK used across 3 business units.
  • Onboarded early users and debugged their deployments until results held up in production.
  • Fine-tuned LLMs for detection tasks and rebuilt ML pipelines as the problem got clearer.

AI/ML Researcher

Sep 2023 - Jun 2024
QuillAudits
  • Built and launched QuillShield, an AI smart-contract auditing tool, as its lead AI engineer.
  • Ran statistical modeling across 100K+ smart contracts for vulnerability research.
  • Turned research into tools, including due diligence analysis for market decisions.
  • Delivered with cross-functional Agile teams and wrote the internal docs that kept results reproducible.

Selected work

Showing 8 of 8 projects

Featured - FinTech / Agents

Financial Strategy Risk Scoring Engine

Crash recall on dated backtests leads: deterministic scorecard over 7 risk axes on point-in-time market data, paired with a groundedness-gated LLM whose verifier rejects any claim it cannot trace to data.

Validated on historical crashes

TypeScriptLLMVerifier

Featured - Security / Agents

Agentic-Auditor

Measured finding precision on labeled contracts: on-the-go AI smart-contract auditor with a Gemini assisted flow and client ready reports in a Vite plus TypeScript front end.

AI assisted audit UX

TypeScriptViteGemini

Security / Fraud response

GoldenHour

Triage checklist for cyber fraud response. Collects evidence fast and drafts the first report.

  • Evidence checklist
  • Draft first report
TypeScript

AI safety

Guardrails

Input/output firewall for LLM apps: prompt-injection and PII checks that run in CI before deploy.

  • Prompt injection checks
  • PII redaction
Python

Voice RAG / Hindi

DesiRAG

Hindi voice RAG over multilingual search passages, with hybrid retrieval and guardrails. Built for HH Goa 2026.

  • Hybrid retrieval
  • Guardrailed
PythonRAGVoice

Security / ML

Vuln-Detector

Experiment harness for contract-vulnerability baselines feeding the production detection models.

  • Baseline models
PythonJupyter

Construction tech / CV

MEP Industrial Elements Detection

Open-set detection of mechanical, electrical and plumbing elements with GroundingDINO, served through a real-time AWS inference API.

  • No fixed label set: open vocabulary
  • Live AWS inference API
GroundingDINOPyTorchAWSPrivate / work project

Public infrastructure / CV

Real-Time Pothole Detection

YOLOv8 detector streaming through Kafka into a Streamlit dashboard, with REST APIs on AWS.

  • Kafka streaming pipeline
  • Streamlit dashboard plus REST APIs
YOLOv8KafkaStreamlitAWSPrivate / work project

Stack

PyTorch and fine-tuned LLMs up front, AWS and vLLM behind them, TypeScript holding the SDK together.

Black chips are daily drivers.

AI / ML
PyTorchTensorFlowTransformersOpen agent frameworksDevtoolsLLM fine-tuning
Computer Vision
YOLOv8GroundingDINOOpenCVVLMs
MLOps & Cloud
AWSOCIDockerCI/CDModel monitoringInference optimizationvLLM
Engineering
PythonTypeScriptBunSoliditySQL/NoSQLREST APIsTDDAgent harness designSDK designEval and red teaming
Data
PandasNumPyMatplotlibPlotlyStreamlit

Certifications

OCI Generative AI Certified Professional

Oracle - Jul 2024

Agentic AI Certified Associate

Oracle - Jul 2026

Copado Certified AI

Copado - Jul 2026

Prompt & AI Engineering Certification

MFT - Jun 2024

Domains

Blockchain / Web3 securityConstruction tech (MEP)Public infrastructure computer visionFinTech risk scoringMulti-modal AI systems

Agents that tell the truth.

Contact

Let's build agents that tell the truth.

Available for remote work, worldwide. Based in Bangalore (IST, UTC+5:30) - replies within a day.

Book a callDownload CV
inghtgmail