Skip to content
ARCHYTASAI

ENV-FIN-104 · Portfolio Management & Investment Compliance

Portfolio Risk & Mandate. Make portfolio decisions within risk, compliance, liquidity and authorization limits.

Simulated portfolio accounting, market data, risk analytics and order management, where agents interpret investment mandates, measure exposures, evaluate proposed trades and keep the portfolio within its restrictions.

Custom buildPortfolio Construction · Risk Analytics · Investment Compliance

Overview. What the environment trains, and the work it's built around.

Core competency

Make portfolio decisions within risk, compliance, liquidity and authorization limits.

  • Turning mandate language into testable restrictions
  • Reconstructing exposures and risk, including look-through holdings
  • Constructing rebalances within concentration, liquidity and cost limits
  • Diagnosing active and passive breaches and proposing remedies
  • Getting approval before trades that need it

Built around the work of

Portfolio Manager · Risk Analyst · Investment Compliance Officer

Simulated systems

  • Portfolio management system
  • Risk engine
  • Order management system
  • Compliance service

Sample task

You manage a simulated long-only equity portfolio. The portfolio has received a new cash allocation, and the investment committee has requested changes to several sector exposures. The mandate contains issuer concentration limits, sector constraints, restricted securities and liquidity requirements. Evaluate the requested changes, construct a compliant rebalance proposal, and document any trade that cannot be executed within the mandate.

Specification. The environment, field by field.

spec · env/portfolio-mandate.yamlENV-FIN-104
product_type:
Executable RL training and evaluation environment
category:
Portfolio Construction · Risk Analytics · Investment Compliance
commercial_model:
Environment development and validated task packages
environment_id:
ENV-FIN-104
target_roles:
Portfolio Manager · Risk Analyst · Investment Compliance Officer
core_competencies:
  • Turning mandate language into testable restrictions
  • Reconstructing exposures and risk, including look-through holdings
  • Constructing rebalances within concentration, liquidity and cost limits
  • Diagnosing active and passive breaches and proposing remedies
  • Getting approval before trades that need it
data:
Synthetic securities, issuers, prices, volumes, benchmarks and fund holdings. Licensed market data can be added where an engagement allows it.
simulated_systems:
Portfolio management system · Risk engine · Order management system · Compliance service
interfaces:
Structured portfolio, risk and order tool API · Mandate document inspection · Python for calculations
interaction:
Stateful, multistep tool use with trade simulation and pre-trade compliance checks, and a simulated investment committee for exception approvals.
state:
Positions, cash, prices, benchmark weights, mandate restrictions, orders and approvals. Each episode starts from an isolated, seeded snapshot.
disturbances:
Ambiguous mandate wording, ETF holdings that hide exposure, thin liquidity, restricted securities, market moves that cause passive breaches and requests that conflict with the mandate.
task_families:
  • TASK-FIN-104-01 · Investment mandate interpretation
  • TASK-FIN-104-02 · Portfolio exposure & risk reconstruction
  • TASK-FIN-104-03 · Constrained portfolio rebalancing
  • TASK-FIN-104-04 · Mandate breach investigation & resolution
evaluation:
Mandate interpretation, exposures, limits and the final portfolio state, all recomputed by the verifier. Compliance is checked before any optimization objective.
reward:
Compliance and authorization gate the reward. Valid proposals are then scored on a stated objective such as tracking error, turnover or cost.
training_use:
RL training, held-out evaluation and SFT on reference proposals.
deployment:
Containerized, isolated execution, with a customer-controlled deployment option.
licensing:
Commercial use rights. Exclusivity, ownership and any third-party data licenses are agreed for each engagement.

Workflow coverage. What agents do, and what the verifier checks.

  • Workflow

    Mandate interpretation

    Example tasks

    Decide which proposed purchases the mandate permits.

    What is verified

    Each decision follows from the mandate's rules and the current holdings.

  • Workflow

    Exposure and risk

    Example tasks

    Compute sector exposure with look-through and active weights.

    What is verified

    Exposures recompute from holdings, fund constituents and benchmark weights.

  • Workflow

    Rebalancing

    Example tasks

    Invest new cash within concentration, liquidity and restricted lists.

    What is verified

    The proposal is compliant at every step, and executable within liquidity limits.

  • Workflow

    Breach handling

    Example tasks

    Classify a breach as active or passive and propose a remedy.

    What is verified

    Classification, remedy and escalation follow the mandate and policy.

Task families. Each family has its own datasheet: prompt, verification criteria and failure modes.

Agent tools. What the agent can call.

The agent reads the mandate and holdings, computes exposures, simulates trades through pre-trade compliance, and submits proposals or exception requests.

tools · env/portfolio-mandateJSON-schema tools
  • get_portfolio
  • get_positions
  • get_market_prices
  • get_investment_mandate
  • calculate_exposures
  • simulate_trade
  • check_pretrade_compliance
  • submit_rebalance_proposal
  • request_exception_approval

Verification methodology. Checked against state the verifier computes itself.

Mandate interpretation, exposure calculations, treatment of cash, currencies and positions, issuer, sector, concentration and restricted-security limits, liquidity and cost assumptions, trade approvals, and the final portfolio state.

The verifier checks

  1. 01Mandate restrictions are interpreted as written.
  2. 02Exposures recompute from holdings, constituents, cash and currency positions.
  3. 03Issuer, sector, concentration and restricted-security limits hold at every step.
  4. 04Trades fit liquidity limits, and cost assumptions follow the model.
  5. 05Trades that need approval have it before execution.
  6. 06The final portfolio state matches the proposal.
  7. 07Valid proposals are scored on the stated objective only after compliance passes.

Reward

Compliance is checked before optimization. Several rebalances can be valid, scored on a stated objective such as tracking error, turnover or cost. Speculative profits are never rewarded.

Scores zero on

  • A proposal that breaches a mandate limit
  • Trading a restricted security
  • Executing a trade that needs approval without it

Included deliverables. What you receive, and what we agree before we build.

Delivered with the environment

  • Containerized environment with seeded portfolios, market data and mandates
  • Validated task instances with reference proposals
  • Verifier code and objective functions
  • Agent traces and failure-mode analysis across repeated runs
  • Technical datasheet and integration documentation

Agreed before development

  • Mandate types and restriction sets
  • Execution, liquidity and cost limits
  • The objective used to compare valid proposals
  • Tolerance thresholds
  • Any third-party data licenses

Every environment is built to order. Tell us which workflow you need to train or evaluate, and where you are in the process.

Request sample tasksDiscuss custom developmentRequest technical integration detailsSee ways to buy