AI & Software Engineer · Baku, Azerbaijan
Riad Mukhtarov
develops software and AI systems in Rust and Python, with a focus on distributed systems and complex infrastructure. Before that, he taught computer science at State University of New York at Buffalo. He likes complicated challenges and clean abstractions.
Education
State University of New York at Buffalo — BS in Computer Science
- Awards: Undergraduate Teaching Assistant Award, Multiple Dean's List Honors
Experience
- Coordinating website renewal, internal platform improvements, and new technology solutions to support organizational operations
- Ensured 99.9% uptime for IT infrastructure supporting 500+ delegates under strict SLAs
- Resolved real-time infrastructure and hybrid session incidents under strict availability requirements
- Assisted with access control, device hardening, and secure system handling
- Collaborated with vendors and technical teams to ensure uninterrupted operations
- Taught core data structures and algorithms in Scala and Java to ~1000 students
- Managed 20 TAs and coordinated course logistics with faculty
- Oversaw grading with emphasis on academic integrity
- Provided weekly reviews and debugging support to ~200 students
- Built backend systems using Python and MongoDB with ADA-compliant frontend components
- Developed authentication and task-based point allocation system
- Optimized WebSockets, cutting server load 90% and enabling instant sync
Projects
Rust
Ollama
fastembedCLI that takes a plain-English dev-tool idea, searches open-source registries for existing implementations, ranks matches with local embeddings, and writes a scoped verdict via LLM.
- — stars on GitHub — downloads on crates.io
- Built concurrent source adapters for crates.io, npm, PyPI, GitHub, Go, Maven, NuGet, RubyGems, Docker Hub, VS Code Marketplace, and Hacker News
- Designed integrity-scoped verdict system that never asserts absence — only reports what was found in sources checked
- Added interactive ratatui TUI with detail view, browser open, and structured JSON output for CI pipelines
Multi-retriever RAG with quorum filtering requiring cross-retriever consensus before surfacing evidence. Achieved 95% recall vs 70% baseline.
- Published to crates.io as an installable Rust crate (cargo add quorumrag)
- Built multi-retriever ensemble with Reciprocal Rank Fusion scoring and quorum filtering
- Implemented overlapping chunk processing with 50% stride to prevent boundary-split answers
- Added parallel embedding with caching to optimize cold-start performance
- Clustered evidence via cosine-similarity centroids so the quorum vote operates on deduplicated semantic groups
almostAykhan — ABB Bank RAG Chatbot
Python
FastAPI
OpenAI
FAISS
DockerRAG chatbot answering questions exclusively from ABB Bank's public content with multilingual support and strict context-only guardrails.
- Built a two-service RAG pipeline (API gateway + QA service) scraping, chunking, and indexing ABB Bank pages via FAISS
- Implemented multilingual support across Azerbaijani, English, and Russian with prompt injection blocking
- Applied distance-gated out-of-scope detection to prevent hallucination — returns 'Bunu bilmirəm' for off-topic queries
- Added SQLite observability and Chart.js query analytics dashboard for monitoring chatbot usage
MatchSentinel — Transaction Monitoring Platform
Java
Spring Boot
RabbitMQ
PostgreSQLEnd-to-end transaction monitoring pipeline with event-driven scoring, async processing via RabbitMQ, and idempotent data handling.
- Designed event-driven architecture for scoring, notifications, and reporting
- Built async processing with RabbitMQ and safe reprocessing
- Implemented per-service databases with Liquibase migrations and environment-based configuration
- Deployed to AWS EC2 with Docker Compose, Linux networking, and public endpoints
Raft-Based Consensus Algorithm
Go
UDP
ProtobufRaft consensus preserving consistency under node failures via leader election and log replication.
- Implemented leader election with randomized timeouts and term-based logical clocks to prevent split votes
- Built log replication via AppendEntries RPC with consistency checks and quorum-based commitment
- Handled node failures and network partitions while tolerating up to ⌊n/2⌋ server failures
- Enforced safety guarantees ensuring elected leaders always contain all previously committed entries
- Designed custom RPC framework over UDP with Protobuf serialization, retransmission, and message deduplication
Distributed Hash Table (Kademlia)
Kademlia DHT for resilient peer discovery and O(log n) lookups in dynamic networks.
- Implemented XOR-based distance metric with k-bucket routing tables for peer organization
- Built iterative node lookup with α concurrent queries refining toward target keys in O(log n) hops
- Designed join and bootstrap protocol for routing table population from seed nodes
- Added key-value republishing and redundant storage across k closest nodes for fault tolerance
Redis Redesign
RedisTwo-layer TTL-aware cache eliminating stale-data inconsistency while preserving low-latency access.
- Replaced a monolithic cache with two layers: key-subkey to id mapping and id to value storage
- Applied TTL lifecycle control to expire stale entries predictably
- Implemented garbage collection to remove expired entries and dangling cross-layer references
- Preserved low-latency access while improving consistency in distributed workloads
Technical Skills
Programming Languages
CBackend & Runtimes
FlaskAI, ML & Agentic Systems
LangGraph
NVIDIA OpenShellCommunication Protocols & APIs
gRPC
RESTful APIs
GraphQL
WebSockets
WebRTCMessaging & Integration
Apache KafkaDatabases & Persistence
MySQL
SQLite
MongoDBCloud, Infrastructure & DevOps
AWS
Supabase
Kubernetes
GitHub Actions
GitLab CI/CD
LinuxTesting & Quality Assurance
Pytest
Selenium
PostmanTeaching
Data Structures
Scala
GitWeb Applications
Software Quality in Practice
TrelloDiscrete Structures
Introduction to Computer Science
Computer Organization
Recommendations
Put him into a Project Manager/Team Lead role, and I guarantee to you that he will shine.
He not only demonstrated a great technical understanding of the content but was able to explain these concepts in a simple manner to students who were struggling to apply the material.
He was very responsive to requests, professional in all his interactions with students and course staff, and technically competent in all the tools and processes covered in the course.
I was impressed by his strong work ethic, positive attitude, professionalism, and dedication. He proved himself to be an organized, efficient, and hardworking TA.
Riad is one of the best individuals I have ever had the opportunity to work with. His ability to communicate complex concepts to a range of skill levels continually impressed me.
I attest to Riad's work ethics, skill, performance, and communication. I do not hesitate to recommend Riad to any employer.
Let's talk.
