Federico Nemirovsky
Software Developer
4+ years ML exp
C++
SQL
C
Python
Rust
System Architecture: 5/10
API Design: 4/10
Scalability & Performance: 4/10
Active 1 day ago
+54 (91) 138322939
Invite to interview
Message
Download CVCV
Overview
Technical skills
Roles
Overview
Distributed systems engineer (senior-level) focused on building concurrent actor-based services with a strength in designing custom leader-election and token-ring coordination protocols. The strongest proven skill is distributed coordination and concurrency as implemented in station/src/token_ring.rs and the coordinator/pending-request handling in station/src/coordinator.rs. Public code does not show persistent storage patterns, production-level authentication for network messages, API versioning or cloud deployment/observability tooling.
Technical skills
C++
SQL
C
Python• Middle • 3y+
Rust• Senior
C++
CMake
PyTorch C++
TensorFlow C++
Rust
Rayon
Actix Web
Tokio
Serde
Databases
Neo4j
RabbitMQ
AI/ML
AI Agents
Anomaly Detection
Function Calling
LangGraph
LightGBM
LlamaIndex
RAG
smolagents
Time Series Forecasting
Transformers
XGBoost
LangChain
Pandas
NumPy
Hyperopt
Stable-Baselines3
Copilot• 4y+
Cursor
Fine-tuning
LLM
PyTorch
Scikit-learn
Spark
TensorFlow
DevOps
Docker
Docker Compose
Git
CI/CD
Senior Backend Developer
Confidence: Medium Distributed Systems
Distributed systems engineer (senior-level) focused on building concurrent actor-based services with a strength in designing custom leader-election and token-ring coordination protocols. The strongest proven skill is distributed coordination and concurrency as implemented in station/src/token_ring.rs and the coordinator/pending-request handling in station/src/coordinator.rs. Public code does not show persistent storage patterns, production-level authentication for network messages, API versioning or cloud deployment/observability tooling.
API Design
4/10
How well APIs are designed
API surface uses custom binary+JSON network messages and actor message types with pragmatic responder patterns, but lacks formal versioning, documented error contract or idempotency keys.
Evidence
fedenemi/distributed-fuel-control-system/station/src/main.rs: multi-message deserialization of NetworkMessage and routing to actor messages
fedenemi/distributed-fuel-control-system/station/src/coordinator.rs: PendingRequestResponder implementations (DashboardResponder, StationResponder) and CoordinatorTcpClient
fedenemi/distributed-fuel-control-system/common/src/messages.rs: (message types referenced across actors)
Data Layer & Database
3/10
Working with databases
Data is in-memory with careful parsing and validation and a change-log to avoid double counting, but no persistent database, migrations or strong transaction boundaries are present.
Evidence
fedenemi/distributed-fuel-control-system/station/src/change_log.rs: ChangeLog.apply_for_port, ParsedOperation parsing/validation and tests
fedenemi/distributed-fuel-control-system/station/src/coordinator.rs: account_info_snapshot and coordinator_accounts usage
fedenemi/distributed-fuel-control-system/station/src/coordinator.rs: initialize_coordinator_accounts reading initial_data.json
Scalability & Performance
4/10
Handling load and speed
Concurrency and scaling patterns are present: actor model with Tokio, explicit threadpool for fork-join processing and token-ring distribution with timeouts, but there is no evidence of load testing, caching/invalidation strategies or production rate limiting.
Evidence
fedenemi/Riesgo-Crediticio-Concurrente-American-Express-Dataset/src/processor.rs: rayon ThreadPoolBuilder and parallel file processing (fork-join)
fedenemi/distributed-fuel-control-system/station/src/token_ring.rs: token passing with tokio::time::timeout and non-blocking TCP sends
fedenemi/distributed-fuel-control-system/pump/src/main.rs: send_to_manager retry loop with delay
System Architecture
5/10
Overall system structure
Clear multi-crate workspace and actor-based decomposition with distinct roles (pump, station, pump_manager, dashboard) and a designed leader-election + token-ring coordination protocol, indicating deliberate system architecture decisions.
Evidence
fedenemi/distributed-fuel-control-system/Cargo.toml: workspace with car, common, dashboard, station, pump, message_sender, pump_manager
fedenemi/distributed-fuel-control-system/station/src/token_ring.rs: leader integration and token lifecycle logic
fedenemi/distributed-fuel-control-system/pump/src/main.rs and station/src/main.rs: separation of responsibilities and network glue
Security & Auth
2/10
Protecting data and access
Some input validation and parsing checks exist, but there is no authentication/authorization on network messages, no token lifecycle handling, and no evident secrets hygiene or dependency audit in code.
Evidence
fedenemi/distributed-fuel-control-system/station/src/change_log.rs: ParsedOperation::new and validate_amount enforcing finite numbers
fedenemi/distributed-fuel-control-system/station/src/coordinator.rs: input parsing and explicit error responses (NotFound / Failure)
fedenemi/distributed-fuel-control-system/station/src/main.rs: raw TCP message handling without authentication or TLS
Reliability & Observability
4/10
Stability and monitoring
Reliability patterns exist: retries with backoff, explicit timeouts, graceful ctrl-c shutdown handling and structured logging macros, but advanced resilience patterns like circuit breakers, distributed tracing or metrics/alerts are not present or not visible.
Evidence
fedenemi/distributed-fuel-control-system/pump/src/main.rs: send_to_manager retry loop with measured delays and attempt limit
fedenemi/distributed-fuel-control-system/station/src/token_ring.rs: tokio::time::timeout usage when passing tokens to next leader
fedenemi/distributed-fuel-control-system/station/src/main.rs: tokio::signal::ctrl_c handling and spawning per-connection tasks with error logging
Expertise
Rust• Middle
Messaging & Real-time• Middle
Industries
Energy & Utilities• Middle
Financial Services• Middle
Technologies
Rust• Senior
RabbitMQ
Tokio
Actix Web
Serde
Rayon
Recommendations
- Implement authenticated, encrypted transport for inter-node messages (TLS or mutual auth) and document the network API/versioning.
- Add persistent state snapshots and a migration path for coordinator_accounts to move from in-memory to durable storage with transactions.
- Introduce structured metrics and distributed tracing to observe token circulation, leader elections and request latencies in production.
- Expand integration tests and fault-injection scenarios (network partitions, leader flapping) to harden election and recovery behavior.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle AI/ML Engineer
Confidence: Medium Data-centric
Data-centric ML practitioner (middle level) with strong feature engineering and experimental pipelines for time-series and algorithmic evaluation. The most proven strength is data pipeline and feature engineering for sales forecasting, demonstrated in Prediccion-de-ventas/time_series_analysis.ipynb where lags, target-style encodings and stacked ensembles are implemented and tuned. Public code shows limited production MLOps, no serving/monitoring, and relatively little automated CI or test coverage.
Model Architecture & Training
4/10
How well models are designed and trained
Applied model training and tuning present (XGBoost/LightGBM hyperopt, LSTM training) with reasonable choices but no custom low-level training loops, distributed training or advanced production training pipelines.
Evidence
Prediccion-de-ventas/time_series_analysis.ipynb: XGBoost hyperopt search and model.fit training loop
Prediccion-de-ventas/time_series_analysis.ipynb: LSTM model definition and training (Sequential/LSTM layers, EarlyStopping)
Data Pipeline & Feature Engineering
7/10
How data is prepared for models
Strong, practical data pipeline and feature engineering for time-series forecasting: extensive joins, target/mean encoding, lag features and aggregated statistics designed to avoid leakage.
Evidence
Prediccion-de-ventas/time_series_analysis.ipynb: extensive feature engineering (lags 1,2,3,6; target/mean-style encodings for type/subtype/city)
Innovation_Curves/Innovation_curves.ipynb: multi-source merges and cleaning steps (patents, population, tertiary datasets)
Experimentation & Evaluation
6/10
How results are measured and tested
Solid experimentation and evaluation practice: hyperparameter search (Hyperopt), held-out validation, explicit metrics reporting and ensemble stacking; lacks dedicated experiment tracking (W&B/MLflow) and tightly versioned reproducibility artifacts.
Evidence
Prediccion-de-ventas/time_series_analysis.ipynb: Hyperopt fmin trials usage and recorded validation RMSEs
generar_mediciones_greedy.py: structured experiment functions (experimento_1_equivalencia, experimento_2_comparacion_calidad) with timing and result dumps
generar_mediciones.py: timed measurements (medir_tiempo_pl/medir_tiempo_greedy) and JSON result export
MLOps & Deployment
2/10
How models are shipped to production
Minimal MLOps: model/artifact saving for competition/submission (joblib, save_model, nn.save) is present but no serving, CI/CD, model versioning, or monitoring.
Evidence
Prediccion-de-ventas/time_series_analysis.ipynb: joblib.dump(lr, 'model.pkl') and modelXG/model files saved
Prediccion-de-ventas/time_series_analysis.ipynb: best_lgbm_model.txt saved and nn.save('nn.h5')
generar_mediciones.py / generar_mediciones_greedy.py: JSON dumps of experimental results (resultados_mediciones.json, resultados_greedy_comparacion.json)
Computational Efficiency
3/10
How efficiently computing resources are used
Some performance measurement and timeout handling for algorithmic experiments (timings, alarms) but little evidence of GPU/ memory optimization, batching optimizations for ML, or quantization/profiling.
Evidence
generar_mediciones.py: medir_tiempo_backtracking / medir_tiempo_pl / timeout context manager
generar_mediciones_greedy.py: medir_tiempo_greedy and timed scalability experiments
tp3.py: ejecutar_algoritmo measures and reports per-algorithm timing
Research Depth & Innovation
3/10
Depth of research and new ideas
Algorithmic and empirical work on greedy heuristics (lookahead/minimum-increment) with experimental validation; solid applied algorithm engineering but not novel research contributions or paper-level reproductions.
Evidence
algoritmos/greedy_minimo_incremento.py: greedy_minimo_incremento and greedy_minimo_incremento_con_lookahead implementations with derivation
generar_mediciones_greedy.py: designed experiments for adversarial cases and equivalence checks between greedy variants
Verified artifacts
Expertise
MLOps & Model Lifecycle• Middle
Industries
Commerce• Middle
Data & Analytics• Middle
Technologies
SQL
C++
LangGraph
LangChain
Neo4j
Spark
Docker Compose
LlamaIndex
XGBoost
Fine-tuning
Scikit-learn
Function Calling
AI Agents
LightGBM
smolagents
Transformers
TensorFlow
Pandas
NumPy
Git
PyTorch
Docker
LLM
RAG
CMake
Stable-Baselines3
TensorFlow C++
PyTorch C++
Anomaly Detection
Time Series Forecasting
Hyperopt
Recommendations
- Develop end-to-end time-series forecasting pipelines for e-commerce (feature engineering, hyperopt tuning, ensembling and submission-ready export).
- Implement experiment-driven model improvements and ablation studies (Hyperopt/stacking/validation folds) for medium-scale ML competitions and production models.
- Build algorithmic benchmarking and profiling tooling for combinatorial algorithms (timing, timeouts, adversarial instance generation) and convert those experiments into reproducible scripts.
- Harden MLOps around model serving and lifecycle (packaging models, CI/CD, model versioning, basic drift/metric monitoring) to take competition-quality models into production.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle DevOps Engineer
Confidence: Medium Generalist
Developer focused on algorithmic problem solving and reliable network systems at a Middle level, with practical strengths in protocol implementation. The strongest proven skill is building reliable UDP-based file transfer protocols and error-recovery logic as shown in file_transfer_networks/protocols/selective_repeat.py and file_transfer_networks/protocols/stop_and_wait.py. There is no public evidence of cloud infrastructure, CI/CD pipelines, container orchestration, or IaC practices.
CI/CD Pipelines
1/10
Automated build and deploy
Minimal CI/CD evidence; only local test-run script exists, no reusable workflows, caching, matrix builds, or deploy gates.
Evidence
dynamic-programming-energy-attack/correr_tests.sh: bash test harness used to run Python tests locally
Infrastructure as Code
Managing servers with code
Not evidenced in public code
Containerization & Orchestration
Working with containers
Not evidenced in public code
Observability & Monitoring
2/10
Watching system health
Basic observability via Python logging and recording of transfer metrics and durations; graphs are generated from measurement output but there is no alerting/SLOs/dedicated observability-as-code.
Evidence
file_transfer_networks/protocols/selective_repeat.py: logging of transfer progress, retransmissions and transfer_rate
file_transfer_networks/lib/server.py: setup_logging function and logging of transfer stats
np-complete-balanced-partitioning/generar_graficos.py: generation of plots from measurement JSON results
Reliability & Incident Response
3/10
Keeping systems up
Demonstrated reliability-focused coding: timeouts, retries, graceful EOF handling and retransmission logic in networking code and timeout context managers in experiments; no formal runbooks or incident postmortems.
Evidence
np-complete-balanced-partitioning/generar_mediciones.py: timeout contextmanager using signal.alarm and TimeoutException for experiment timeouts
file_transfer_networks/protocols/selective_repeat.py: retransmission timers, ACK handling, EOF handshake and retry loops
file_transfer_networks/protocols/stop_and_wait.py: retransmission attempts and timeout handling in send/receive loops
Cloud & Cost Optimization
Smart use of the cloud
Not evidenced in public code
Expertise
Service Mesh & Networking• Middle
Observability & Monitoring• Middle
Technologies
Python• Middle • 3y+
Recommendations
- Develop UDP-based or custom transport services and robust network daemons where low-level socket control and retransmission logic matter.
- Implement and benchmark algorithmic components, performance experiments and measurement pipelines (benchmark harnesses, reproducible experiments, plots).
- Harden the transfer system by adding automated tests, CI integration, container packaging, and observability (metrics, traces, alerting).
- Design protocol fuzzing and fault-injection tests to validate reliability under network faults and to convert experiment scripts into CI jobs.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: