Agent skill cards

I curate market data, flag anomalies, and kick off a HybridRAG refresh whenever fresh data lands.

I turn raw data into model-ready features.

I train ML models and forecast returns.

I run multi-step AIQ Deep Research (planner → researcher → synthesizer → citer).

I read 10-Ks, 10-Qs, transcripts.

I read price and volume only.

I read what humans are saying.

I find latent factors and regimes.

I fuse every research view (technical · fundamental · sentiment · ai-factor · predictive · deep-research · hybridrag) and the regime tag into a per-ticker conviction signal. When top-candidate conviction is weak (< 0.30) I auto-kick a DeepResearchRequested on that name so the next cascade has stronger views.

I challenge today's signal and learn from yesterday's. When my critique fails, I auto-fire a BacktestRequested so we know whether the failure is regime-local or systemic.

I run candidate strategies on history; when a run clears the Sharpe + max-DD floor I promote it as a candidate the PM records.

I solve the CVaR problem (cuFOLIO) and blend the result with the active NemoRL policy when one is loaded. MetaAgent's CritiqueFailed gates me — I skip the rebalance and emit RebalanceSkipped.

I make the math executable.

I split the household across sleeves before anyone sizes a trade.

When the PM approves an allocation, I gate it through Compliance and make it the household's truth.

Nothing trades unless it's clean.

I place orders and follow the schedule.

I watch positions during the day.

I tell you what happened — KPIs + IS/VWAP/slippage.

I am the PM. I observe every meaningful event on the A2A bus — research views, signal fusions, compliance verdicts, fills, RL retrains, AutoResearch sessions, HybridRAG retrievals, regime shifts — so when the operator asks me a question I can synthesize across the whole platform. I approve/reject rebalances, auto-fire HybridRAGQuery on >15% concentration risk, halt AutoResearch on hard compliance vetoes, and record candidate strategies.

I launch NVIDIA NeMo-RL training runs (SFT/DPO/PPO/GRPO/DAPO/GDPO/RM/distillation) via the bridge — subprocess into the dedicated Python 3.13 env where nemo-rl 0.6.0 lives.

I count preference pairs from PreferenceLearningAgent (approve/reject on rebalances + narrations). Once enough accumulate, I emit TrainNemoRLRequested(algo='dpo') so the Nemotron policy gets a fresh DPO retrain on the latest user feedback.

I turn every Approve/Override/Reject into a DPO training row.

I write every meaningful bus event to immutable JSONL — full rebalance lifecycle (Proposed → Constructed → Cleared/Blocked → Approved → OrderPlaced → OrderFilled), capital allocator lifecycle, NeMo-RL training lifecycle, critique verdicts, data anomalies. The chain is replayable; nothing the platform emits goes unrecorded.

I tag the cross-asset regime from VIX, 10Y yield, and DXY.

I score symbols by net insider buying from SEC Form 4 in the last 30 days.

I score symbols by call/put open-interest skew on the options chain.

I score symbols by yield × payout sustainability × 5y dividend growth.

I extract a typed knowledge graph from a company's filings and news, fuse it with vector retrieval, and answer multi-hop questions with citations.

Karpathy-pattern meta-loop over NeMo-RL. Each iteration, Nemotron 3 Super proposes a typed config edit (KL penalty, learning rate, batch size); the loop launches the inner DPO/GRPO/SFT run via the NeMo-RL bridge, parses the eval metric, keeps or reverts.