PM.

Pranav Meraga / Engineering portfolio

Robotics &
applied AI.

I’m Pranav, a UC Berkeley aerospace engineer building Cygnus Robotics. My work spans robot data collection and curation, camera systems, and evaluations that make model failures inspectable.

Explore my work
UC Berkeley · Aerospace EngineeringPython / TypeScript / ROS 2 / gRPC

BUILDING CYGNUS ROBOTICS

Training data with a traceable history.

Tools to inspect robot demonstrations, retain recovery behavior, and audit the frames selected for training.

Visit Cygnus Robotics

SELECTED WORK

Robots, data, and model reliability.

Source code, technical decisions, and measured evidence.

Computer vision · 3D measurement

Stereo Truck-Load Volume Estimation

Reconstructed a load in metric 3D, aligned it to the truck bed, and evaluated how occlusion and registration affected the volume estimate.

  • Calibrated stereo
  • 3D registration
  • Spatial interpolation
AI EVALUATION / BIOMEDICAL ANALYTICS
Same data.
Different validation.
0.489Forward monitoring
0.657Random validation
Verified references · Rubric audits · Reproducible evidence
AI AGENT EVALUATIONBiomedical research case study

Benchmark the reasoning behind the answer.

Designed biomedical agent tasks, verified statistical reference solutions, and audited paired responses and scoring rubrics. A longitudinal pain-decoder case shows how validation design changes a research decision.

  • AI benchmarking
  • Biostatistics
  • Temporal validation
  • Python
Computer vision · Real-time systems

Real-time Ball Tracking

A browser receives synthetic video over WebRTC, detects the ball from decoded pixels, and compares its position with timestamp-matched coordinates sent over WebTransport.

  • Go
  • WebRTC
  • WebTransport
  • Computer vision
RETRIEVAL / CANDIDATE SEARCH

Candidate retrieval

RetrieveRe-rankMatch
MACHINE LEARNINGAssessment project

Better search. Better matches.

A candidate search engine combining vector and lexical retrieval, reciprocal rank fusion, cross-encoder re-ranking, and an LLM judge to evaluate role-specific criteria.

  • Python
  • Turbopuffer
  • Voyage AI
  • OpenAI
DEVELOPER TOOLS / CODE INTELLIGENCE
code_intelligence.py
index(apex, metadata)
  ↓
SQLite dependency graph
  ↓
query(symbols, relationships)
  ↓
Compact context for agents
DEVELOPER TOOLSAssessment project

Make a codebase legible to AI.

A Salesforce code-intelligence pipeline built as a machine learning engineering assessment. It indexes Apex and metadata into SQLite, then exposes deterministic, compact queries for LLM agents.

  • Python
  • SQLite
  • MCP
  • 21 tests passing
AGENT SYSTEMS / BENCHMARKING
agentic-harness
Submit taskRun agentGrade + record
Docker-isolated evaluation, one command at a time.
AGENT SYSTEMSCoding-agent evaluation

One harness. Many coding agents.

A portable benchmark runner that queues tasks, runs setup, agent, and grading phases in isolated Docker containers, and stores results for repeatable evaluation.

  • Python
  • Docker
  • SQLite
  • OpenAI Responses API
SYSTEMS / BACKTESTING
Backtest.
Inspect.
Iterate.
Market data → Strategy → Evaluation
SOFTWARE ENGINEERINGAssessment project

A clearer view of a strategy.

Improvements to a supplied futures backtesting system: PnL exception handling, crossed-quote validation, multi-strategy visualization, custom indicators, and unit tests.

  • NautilusTrader
  • Python
  • React
  • Databento
REASONING / ARC-AGI
INPUT
OUTPUT
Learn the pattern. Predict the transformation.
REASONINGReasoning assessment

From examples to a new answer.

An LLM agent for ARC grid puzzles, with structured prompts, text and JSON grid parsing, and an evaluation harness over 20 included puzzles.

  • TypeScript
  • Anthropic
  • OpenAI
ANOMALY DETECTION / MACROBASE
Transaction featuresIsolation ForestAlerts + explanations
Score / Classify / Explain
APPLIED MACHINE LEARNINGAML transaction analysis

Find the unusual. Explain why.

A MacroBase-inspired pipeline that scores transactions, compares percentile thresholds, and explains anomalies with overrepresented categorical attributes.

  • Isolation Forest
  • pandas
  • scikit-learn

RESEARCH

Where it started.

Machine learning meets engineering data.