Skip to content

AI Solutions

AI-powered systems, teams and tools — wired into the work you already do

SakhiSoft is an IT services company that builds AI into production software. Not demos. Systems with evaluation sets, human review where it matters, guardrails and monitoring — the parts that decide whether an AI feature is still working six months after launch.

What we build

Nine AI solutions we deliver end to end

Each one starts from a problem someone in your business is currently solving by hand. If we cannot name that person and that problem, we will tell you it is not ready to build.

AI Copilots & Chat Assistants

Your answers already exist — in Confluence, a policy PDF, six years of tickets, a database nobody queries. Staff still ask a colleague, and the colleague is in a meeting.

What we build

  • Retrieval-augmented assistants grounded in your own corpus, with citations back to the source paragraph
  • Role-aware retrieval so a user only ever retrieves documents their account is already permitted to read
  • Tool-calling copilots that can read from and write to your CRM, ticketing or internal APIs
  • Escalation paths to a human when confidence or retrieval coverage is low
  • Embedded chat surfaces: in-app widget, Slack/Teams bot, or an internal agent console

Typical stack

OpenAI / Anthropic / LlamaLangChainpgvectorQdrantRedisNext.jsFastAPI

Example outcome

Support and ops teams stop paging subject-matter experts for questions the documentation already answers, and every answer carries a link to the clause it came from.

Healthcare & EMSTelecom & ISPStartups & SaaS

Document Intelligence & OCR

Invoices, claims, facesheets and PODs arrive as scans, faxes and phone photos. Someone re-types them into a system, at a few minutes each, with a typo rate nobody measures.

What we build

  • Extraction pipelines that turn PDFs, scans and photos into typed, schema-validated records
  • Layout-aware parsing for tables, multi-page documents and rotated or low-quality scans
  • Per-field confidence scores, so only genuinely uncertain fields go to a person
  • Human-in-the-loop review UI showing the source page with the extracted value highlighted in place
  • Corrections captured as labelled data and fed back into the evaluation set

Typical stack

PythonTesseractAWS TextractGoogle Document AIOpenAI vision modelsPostgreSQLCelery

Example outcome

High-confidence fields post straight through to the system of record; reviewers spend their time on the genuine exceptions rather than on every document.

Healthcare & EMSLogistics & Supply ChainFinTech & Subscription Products

Workflow & Process Automation

A process spans five tools and three humans whose only job is to copy state between them. It works until someone is on leave.

What we build

  • Agentic pipelines that plan a multi-step task, call tools, and stop for approval at the steps that carry real risk
  • Durable orchestration with retries, idempotency keys and dead-letter queues — no silent failures
  • Integrations across CRM, ticketing, billing, telephony and shipping APIs
  • Deterministic rules for anything auditable; models only where judgement is genuinely required
  • A run log for every execution: inputs, tool calls, model outputs, decisions and who approved what

Typical stack

Node.jsPythonTemporalLangGraphBullMQWebhooksPostgreSQL

Example outcome

Routing, triage and data-entry steps run unattended overnight, and the exceptions land in a queue with the context needed to resolve them in one pass.

Logistics & Supply ChainTelecom & ISPHealthcare & EMS

Predictive Analytics & Forecasting

Demand, churn and capacity are forecast in a spreadsheet that one person maintains and everyone quietly distrusts.

What we build

  • Demand, churn, capacity and revenue models trained on your historical data
  • Backtesting against held-out periods, reported honestly against a naive baseline
  • Feature pipelines that are reproducible in training and serving — no train/serve skew
  • Forecasts surfaced where decisions are made: the dashboard, the planning sheet, the API
  • Drift monitoring, with retraining triggered by measured degradation rather than a calendar

Typical stack

Pythonscikit-learnXGBoostProphetpandasdbtPostgreSQLMetabase

Example outcome

Planning moves from monthly gut-feel to a forecast with a published error band that stakeholders can actually reason about.

Logistics & Supply ChainFinTech & Subscription ProductsStartups & SaaS

Computer Vision

Cameras are already recording. Nobody watches the footage until after something has gone wrong.

What we build

  • Detection, classification and counting models for defects, PPE compliance, occupancy and damage
  • Image-quality gates that reject unusable frames before they reach the model
  • Edge or on-prem inference where bandwidth, latency or data residency rules it out of the cloud
  • Annotation workflow and dataset versioning so the training set is an asset, not a one-off
  • Review queues for low-confidence detections, with the frame and the bounding box attached

Typical stack

PyTorchYOLOOpenCVONNX RuntimeTensorFlowNVIDIA Triton

Example outcome

Conditions that previously surfaced in a post-incident review get flagged while the shift is still running.

Logistics & Supply ChainHealthcare & EMS

Recommendation & Personalization

Every user sees the same catalogue, the same dashboard and the same email. Relevance is left to the search box.

What we build

  • Recommenders combining collaborative filtering, content embeddings and business rules
  • Cold-start handling so new users and new inventory are not invisible on day one
  • Real-time feature serving for session-aware ranking
  • Hard constraints layered on top of the model: margin floors, stock, compliance and eligibility
  • A/B testing harness — no ranking change ships without a measured lift

Typical stack

PythonPyTorchPinecone / pgvectorRedisKafkaFastAPI

Example outcome

Ranking changes are judged on conversion and retention measured in an experiment, not on how the demo looked in the review meeting.

Startups & SaaSFinTech & Subscription Products

LLM Integration & Fine-tuning

The team wired up a model, the demo impressed everyone, and then quality, latency and the invoice all became somebody's problem.

What we build

  • Model selection driven by an evaluation set built from your data — not a leaderboard
  • Prompt and context engineering, structured output with schema validation and repair
  • Fine-tuning or LoRA adaptation where prompting has measurably run out of road
  • Cost and latency engineering: caching, routing cheap requests to small models, batching, streaming
  • A provider abstraction so switching or adding a model is a config change, not a rewrite

Typical stack

OpenAIAnthropicLlamaMistralvLLMLoRA / PEFTLiteLLM

Example outcome

The same task runs against a smaller or cheaper model wherever evaluations show no quality loss, and the bill stops scaling linearly with usage.

Startups & SaaSHealthcare & EMSTelecom & ISP

Data Engineering & RAG Platforms

Retrieval quality is a data problem wearing a model costume. Bad chunking and stale indexes produce confident, wrong answers.

What we build

  • Ingestion pipelines from S3, SharePoint, Confluence, databases and email, with incremental sync
  • Chunking and embedding strategies tuned against a retrieval evaluation set, not chosen by default
  • Hybrid search — dense vectors plus BM25 keyword — with a reranking pass
  • Access-control metadata carried through to query time so retrieval cannot leak across tenants or roles
  • Freshness guarantees: deletions propagate, re-embedding runs on change, indexes are versioned

Typical stack

pgvectorQdrantPineconeElasticsearchAirflowdbtPythonS3

Example outcome

Retrieval recall is a number the team tracks on every index change, rather than a quality that gets debated after a user complains.

Healthcare & EMSTelecom & ISPStartups & SaaS

MLOps & AI Governance

The model went to production and the org lost the ability to answer three questions: is it still working, why did it say that, and who signed off.

What we build

  • Evaluation suites run in CI — prompt and model changes are blocked on regression, like any other code
  • Online monitoring: quality proxies, latency, cost per request, refusal and fallback rates
  • Guardrails at both ends — input filtering, PII redaction, output validation, jailbreak resistance
  • Full audit trail: prompt version, model version, retrieved context, output, reviewer, timestamp
  • Rollback paths and kill switches that a non-engineer can operate under pressure

Typical stack

MLflowLangSmithRagasPrometheusGrafanaOpenTelemetryGitHub Actions

Example outcome

A regression is caught by an evaluation run in CI before release, and every production answer can be reconstructed months later for an audit.

Healthcare & EMSFinTech & Subscription ProductsTelecom & ISP

How we deliver

How we deliver AI that survives production

Most AI projects do not fail at the model. They fail because nobody defined correct, nobody measured drift, and nobody could explain an output three months later. This is the checklist we work to.

01

An evaluation set before a single prompt

We start by collecting real inputs and agreeing the correct outputs with your subject-matter experts — typically 50-200 cases including the awkward ones. That set is the definition of done. Without it, 'the AI got better' is an opinion.

02

Offline evaluation gates every change

Prompt edits, model swaps and chunking changes run against the evaluation set in CI and report accuracy, retrieval recall, latency and cost per case. A change that regresses a metric does not merge — the same bar we apply to a failing unit test.

03

Human-in-the-loop where the cost of being wrong is real

Confidence thresholds decide the routing: high-confidence outputs post through, everything else lands in a review queue with the source document and the model's reasoning attached. Reviewer corrections are captured as labelled data and flow back into the evaluation set.

04

Guardrails on both sides of the model

Inbound: PII redaction, injection filtering, and retrieval scoped to the caller's existing permissions. Outbound: schema validation, citation checks against the retrieved context, and refusal when retrieval coverage is too thin to answer honestly.

05

Monitoring that watches quality, not just uptime

A 200 OK from a model API tells you nothing about the answer. We track refusal rates, retrieval-miss rates, output-validation failures, latency percentiles and cost per request, and alert on drift in the distribution of inputs.

06

Auditability by default

Every inference is logged with its prompt version, model version, retrieved chunks, output, and any human decision applied to it. When someone asks in nine months why the system did what it did, the answer is a query, not an investigation.

07

Deterministic code wherever a model is not required

Models are used for judgement, ambiguity and language. Arithmetic, eligibility rules, routing tables and anything an auditor will read stays in ordinary, testable code. This is the single biggest reliability decision on most AI projects.

08

A rollback path you can use at 2am

Model and prompt versions are pinned and deployable independently of application code, behind flags. Reverting a bad prompt should not require a release train, and disabling an AI feature entirely should be a switch anyone on-call can flip.

AI tech stack

The tools we reach for — and why

We are not loyal to a vendor. Model and infrastructure choices are made against your evaluation set, your latency budget and your data-residency rules, then revisited when the numbers change.

Models & providers

Hosted frontier models where quality leads, open-weight models where cost, latency or data residency leads.

OpenAIAnthropic ClaudeLlamaMistralAzure OpenAIAWS Bedrock

Orchestration & agents

Framework for chaining retrieval, tool calls and approval steps into something durable enough to run unattended.

LangChainLangGraphLlamaIndexTemporalModel Context ProtocolCelery

Vector & retrieval

Hybrid retrieval — dense embeddings plus keyword search plus a reranker — sized to the corpus rather than the hype.

pgvectorQdrantPineconeWeaviateElasticsearchCohere Rerank

ML & training

Classical ML and deep learning for the many problems where a fine-tuned small model beats a prompt.

PyTorchTensorFlowscikit-learnXGBoostHugging FaceLoRA / PEFT

Vision & documents

Extraction and detection pipelines for scans, photos, video frames and multi-page PDFs.

OpenCVYOLOTesseractAWS TextractGoogle Document AIONNX Runtime

Data pipelines

The ingestion, transformation and freshness layer that decides whether retrieval works at all.

AirflowdbtKafkaPostgreSQLS3Redis

Evaluation & observability

Evidence that the system still works: regression suites in CI, tracing in production, alerts on drift.

RagasLangSmithMLflowOpenTelemetryPrometheusGrafana

Serving & infrastructure

Where the model actually runs, and how it scales without the bill scaling with it.

FastAPIvLLMNVIDIA TritonDockerKubernetesAWS / GCP / Azure

Industry mapping

Where these solutions land

The engineering is transferable; the domain rules are not. These are the sectors where we already know the workflows, the integrations and the compliance constraints.

Healthcare & EMS

Dispatch systems, patient portals, OCR for facesheets, and HIPAA-aware workflows.

AI solutions that apply

  • AI Copilots & Chat Assistants
  • Document Intelligence & OCR
  • Workflow & Process Automation
  • Computer Vision
  • LLM Integration & Fine-tuning
  • Data Engineering & RAG Platforms
  • MLOps & AI Governance

Logistics & Supply Chain

Shipment tracking, label generation, and carrier integrations (Shippo, DHL, UPS, etc.).

AI solutions that apply

  • Document Intelligence & OCR
  • Workflow & Process Automation
  • Predictive Analytics & Forecasting
  • Computer Vision

Telecom & ISP

Customer portals, recurring billing, UISP/UNMS integrations, and usage dashboards.

AI solutions that apply

  • AI Copilots & Chat Assistants
  • Workflow & Process Automation
  • LLM Integration & Fine-tuning
  • Data Engineering & RAG Platforms
  • MLOps & AI Governance

Startups & SaaS

From MVP to scale-up: rapid prototypes, investor-ready demos, and production hardening.

AI solutions that apply

  • AI Copilots & Chat Assistants
  • Predictive Analytics & Forecasting
  • Recommendation & Personalization
  • LLM Integration & Fine-tuning
  • Data Engineering & RAG Platforms

FinTech & Subscription Products

PCI-conscious payment flows, invoicing, and recurring revenue analytics.

AI solutions that apply

  • Document Intelligence & OCR
  • Predictive Analytics & Forecasting
  • Recommendation & Personalization
  • MLOps & AI Governance

Qualifier

Is your use case actually a fit?

We would rather lose the project than build something that quietly stops working. Read both columns honestly — the right-hand one is the one that saves you money.

Worth building

  • The task is repetitive, language- or document-heavy, and a competent person could do it with the information already available to the system.
  • You can produce examples of the work being done correctly — past tickets, reviewed documents, historical decisions. That is your evaluation set and your training data.
  • A wrong answer is recoverable: it gets caught by review, a rule, or a human before it reaches a customer or a ledger.
  • There is an owner on your side who can arbitrate what 'correct' means when two reviewers disagree.
  • Success has a number attached — handling time, straight-through rate, deflection rate, forecast error — that someone already tracks.

We will tell you not to

  • The rules are fixed, written down and unambiguous. That is a deterministic system with a test suite, and it will be cheaper, faster and auditable. We will tell you so.
  • A single wrong output is unrecoverable — money moves, a dose is administered, a contract executes — with no review step you are willing to fund.
  • The underlying data does not exist, is not accessible, or is so inconsistent that your own experts cannot agree on the right answer.
  • The goal is to be able to say the product has AI in it. That is a marketing brief, and it will not survive contact with production.
  • Accuracy needs to be effectively perfect on unbounded natural-language input. Nothing available today clears that bar honestly.

Still unsure? That is the normal state. A short discovery call is usually enough to tell whether your problem needs a model, a rules engine, or just a better form.

Bring us the workflow, not the buzzword

Tell us which task is eating your team's week, what “done correctly” looks like, and what data you already have. We will come back with a scoped first phase, an evaluation plan, and an honest read on whether AI is the right tool for it at all.