Born a poet, works as an engineer

Feyaz Baker  • Peerlist

LLM-driven SQL Generation and Analysis System using LangGraph, FastAPI, PostgreSQL, FAISS • Implemented pipelines for a system that converts natural language queries into executable SQL over multi-table PostgreSQL datasets, supporting complex analytical queries such as aggregations, joins, and ranking. • Coordinated agent-based workflows using LangGraph, enabling iterative query refinement, conditional execution, and failure recovery, reducing invalid query generation by 40%. • Devised a lightweight tool server (MCP-style API) using FastAPI to expose database operations for improved modularity. • Integrated a retrieval layer using FAISS and sentence-transformer embeddings to store and reuse past successful queries, improving SQL generation accuracy over time. • Added a human feedback loop to capture correctness signals, storing successful queries and tracking failed patterns to guide future work. • Enforced schema-aware SQL generation with strict constraints and encouraged use of window functions for efficient analytical queries

Autonomous Data Analyst

A Web portal for one of the field offices of Comptroller and Auditor General of India . The portal is basically an interface for the office of about 40 employees to organize and manage their data with complex relations with the pwer of a relational database management system . The tools and technologies used in this project are Django , HTML ,CSS , JavaScript and MySQL. The codebase is official so i have only shared a link to the letter of appreciation from the office we built the portal for.

MIS Web Portal (Management Information System

Pioneered a new ANTLR-based grammar to represent data flow transformations in a human-readable format. Engineered parsing and execution layers in Java, supporting transformation rules across distributed Spark pipelines. Maintained backward compatibility with legacy HRF formats, enabling seamless migration without disrupting existing workflows. Delivered 150+ features and enhancements based on production feedback, improving system reliability and developer adoption. Reduced average compile time significantly through parser and execution optimizations. Incorporated multiple customer-driven features, aligning system capabilities with real-world data transformation needs.

Locode Development (DAS Script Tool)

Configured a GenAI synthetic data pipeline using OCI AI APIs generating up to 2M semantically coherent rows/min for large-scale ML training datasets. Generated pattern-consistent synthetic datasets with realistic temporal drift and cluster structures for ML model training and validation. Demonstrated enterprise impact through VP-level technical presentation, validating GenAI-driven data generation for production analytics workflows. Deploying the platform across three engineering teams in India, integrating GenAI data generation into production ML and analytics pipelines.

Data Augmentation Suite (GenAI Synthetic Data

Built an interactive visual debugging layer during Orahacks 2024, reducing pipeline debugging time by ~40–60% by enabling real-time inspection of table hierarchies and custom data flow orchestration. Integrated with existing LoCode using ANTLR-based parsing and Spark execution, supporting 20+ pipeline configurations with minimal overhead. Scaled the prototype into a production-grade internal tool, improving developer productivity and significantly reducing onboarding time for new users. Deployed as a production companion tool across customer dashboards, contributing to increased adoption of LoCode pipelines.

Visual Debugger

Implemented a student-teacher architecture in PyTorch with consistency-based regularization on ACDC, Left Atrium, and ISIC 2018 datasets. Achieved consistent performance improvements over existing state-of-the-art methods, with results published as part of a major academic project.

SSL Medical Image Segmentation

Created and deployed a full-stack Django system for audit information automation. Received a Letter of Appreciation from the Office of the Principal Director of Audit (Central), Bangalore.

IA&AS Digitization

Developed a computer-aided simulation framework for a complex, unsolved variant of Ultimate Tic Tac Toe, analyzing multiple game strategies and deriving win/draw probabilities through data-driven methods. Presented the work at NIT Jalandhar and received the Best Paper award at the conference session.

An Analysis of Heuristics for a mathematicall

Hackathon aimed at improving college life. Built a keyless card entry authentication system with Texas Instruments hardware.

Imagineer Hackathon - Keycard Entry system

Developed a small POC model network implementing Paillier Homomorphic Encryption libraries, enabling secure AI modelling on completely obscured and encrypted data.

Paillier Homomorphic Encryption Based Neural 

Demonstrated analytical approaches to improve medical inventory prediction using a Python-based streaming data framework. Presented this work at IEEE TENSYMP 2021, held in Jeju-do, South Korea.

Parallel Data Systems with Self-Generating An

Selected among a small cohort of applicants to engineer client-based technology solutions. Worked with Raspberry Pi and microcomputers for wireless monitoring and maintenance systems. Improved user interfaces and resolved connectivity issues to enhance usability. Developed an OS package for remote deployment via secure FTP.

Remote Software Deployment Tool

Improved TRP solution using GRASP-VNS with CUDA optimization, implementing and validating results on a computation cluster while gaining experience in high-performance computing and Linux systems. Achieved significant error reduction by adapting sequential optimization algorithms for parallel execution.