About
I am a software engineer who builds systems from the ground up - from programming languages and compilers to distributed AI infrastructure. Currently leading technical teams at IHairium, I specialize in high-performance distributed systems, low-latency applications, scalable AI inference, and complex platform architectures.
Featured Projects
Eridu: A C-Like Programming Language
- Developed a C-like statically-typed programming language from scratch in C, supporting first-class functions, abstract data types, and multi-paradigm constructs.
- Built custom code generation system, directly outputting x86-64 AT&T assembly with custom intermediate representation, stack frame management, and memory allocation.
- Built a custom testing framework in Python with a domain-specific language for inline test specification, enabling test-driven development by validating expected outputs of code snippets.
August: Proof of Work Blockchain Implementation
- Implemented a Bitcoin-inspired blockchain from scratch in Go, featuring proof-of-work consensus, longest-chain fork resolution, and gas-metered smart contract execution on a custom stack-based virtual machine (AVM) based on SHA-256 and Ed25519 signatures.
- Designed Marigold, a statically-typed smart contract programming language with complete compiler pipeline (lexer, parser, type checker) that targets August Virtual Machine Bytecode (AVMBC), alongside the custom virtual machine featuring 35+ opcodes with 256-bit arithmetic, persistent storage, and blockchain context instructions.
- Built full peer-to-peer networking layer with decentralized peer discovery and headers-first chain synchronization, alongside comprehensive toolchain including wallet, miner, and HTTP API and tests for multi-node network validation.
Tribune: Distributed Multi-Party Computation Library
- Built a modular C++ framework for secure multi-party computation (SMPC) with plug-in interfaces for cryptographic protocols and data-sharding strategies.
- Engineered a peer-to-peer producer–consumer protocol where clients exchange masked data shards, while a coordinating server re-aggregates them to reveal only the global result-preserving privacy of individual inputs.
- Demonstrated the framework by training a federated machine-learning model that performed linear regression on device usage patterns, enabling predictive analytics without exposing personal routines or raw client data.
Work Experience
Technical Lead (Contract)
- Hired to lead a targeted effort transitioning the platform from B2C to B2B, building scalable AI infrastructure and multi-tenant support.
- Extended Java 17+ Spring Boot backend to support multi-tenant B2B functionality, allowing partners to define custom diagnostic flows, recommend their products, and collect per-tenant metrics.
- Built scalable Python microservices (Flask, Docker, AWS EC2) replacing the legacy diagnostics engine, delivering low-latency inference services, complete with CI/CD and full test coverage.
- Built an AWS SQS-based producer–consumer queue to scale AI services horizontally and improve fault tolerance. Additionally, reduced inference latency by 40% through thread pooling, bottleneck profiling, memory optimization, and runtime upgrades.
- Managed a cross-functional team of three engineers across iOS, frontend, and backend, overseeing Agile workflows including sprint planning, task estimation, and code reviews.
- Coordinated with external AI developers: validating results, writing delivery specifications, generating reports, and integrating their AI models into the platform.
Backend Engineer
- Reduced response times by 88% on high-load FastAPI endpoints through PostgreSQL query optimization, payload minimization, caching, and asynchronous task offloading with Celery - enabling efficient queuing and scalable worker orchestration.
- Built a fully automated MLOps pipeline in collaboration with AI specialists to deploy multi-tenant inference models on Azure Kubernetes, integrating canary and blue-green rollouts with build-metadata for traceability, and automating model warehousing in blob storage.
- Engineered a configurable ETL system and built a rule-based sanitation engine, converting raw security logs into compliance-aligned datasets (GDPR, NIS-2) for LLM-powered batch inference and structured audit reporting.
- Managed dozens of production rollouts across customer environments, managing infrastructure and schema migrations with Alembic, Docker, and Kubernetes; integrated CI/CD pipelines with PyTest coverage enforcement.
Freelance Software Engineer
- Developed and launched a cross-platform Flutter app for a marketing client in just 3 weeks, achieving 1,500+ downloads in the first week and generating 500k+ social media impressions.
- Developed a Python/FastAPI WhatsApp bot with Recruitee and WhatsApp Cloud API integrations to automate post-sales surveys, enabling cross-department analytics, survey completion rates, question drop-offs, and engagement trends.
- Developed JavaScript/Vue.js websites for half a dozen companies, integrating with CRM systems, booking systems, and third-party APIs to streamline client operations and improve user experience.
Software Engineering Intern
- Developed kernel-level and embedded systems in C/C++ for robotics platforms, contributing domain knowledge to support predictive model feature engineering using PyTorch
Technical Consultant
- Created a Python microservice for GDPR-compliant data sanitation, processing 130M+ emails with automated testing and CI/CD pipeline.
Research
Research Scientist
September 2023 to June 2024
- Developed OrthAdam, a novel Adam optimizer variant that achieved ~2% accuracy improvements compared to state-of-the-art AdamW and AdamP optimizers on CIFAR-100, CIFAR-10, and SVHN datasets across ResNet18, DenseNet121, and ResNeXt50 architectures.
- Created AttentionSplit, a hybrid neural network layer combining recurrence and attention mechanisms for temporal data analysis, achieving marginally better accuracies than state-of-the-art LSTM and Transformer models on Acrobot-v1, CartPole-v1, and MountainCar-v0 environments - particularly promising as a first iteration of the research.
- Conducted comprehensive evaluation across computer vision benchmarks (ResNet18, DenseNet121, ResNeXt50) and reinforcement learning environments (Acrobot, CartPole, MountainCar, HalfCheetah, Walker2D) to validate architectural innovations.
- Published research findings on the importance of scale-invariant momentum-based algorithms and the potential of attention-recurrence combinations in deep learning applications.
Education
Master of Science in Computer Science
Microservices & Dev(Sec)Ops, Software Engineering & Agile, Computer Vision, Artificial Intelligence
Bachelor of Science in Computer Science
Data Mining & Machine Learning, Compiler Construction & Optimization, Database Systems, Generic Programming (C++)